Michael/FlowReader
1
0 Fork 0
mirror of https://github.com/R0m1k3/FlowReader.git synced 2026-10-14 02:39:33 +02:00
Code Issues Packages Projects Releases Wiki Activity

feat(ux): fix focus swipe animation, button visibility and mobile scroll

Browse Source
This commit is contained in:
Michael committed 2026-02-13 08:35:13 +01:00
1 parent 22e002f47f
commit 9d41a74b36
466 files changed
+6111 -42082
Download Patch File Download Diff File Expand all files Collapse all files

No files matched your search

_bmad/bmb/workflows/agent/data/agent-architecture.md
+258
View File
@@ -0,0 +1,258 @@
# Agent Architecture
Single Agent type with `hasSidecar` boolean. `critical_actions` decoupled from sidecar.
## Decision Matrix: hasSidecar
| hasSidecar | Structure | Use When |
|------------|-----------|----------|
| `false` | Single YAML file (~250 lines) | Stateless, single-purpose, personality-driven |
| `true` | YAML + sidecar folder | Persistent memory, long-term tracking, relationship-driven |
---
## YAML Schema
```yaml
agent:
metadata:
id: _bmad/agents/{agent-name}/{agent-name}.md
name: 'Persona Name'
title: 'Agent Title'
icon: '<emoji>'
module: stand-alone # or bmm, cis, bmgd
persona:
role: | # First-person, 1-2 sentences
identity: | # Background, 2-5 sentences
communication_style: | # Voice, tone, mannerisms
principles: # Core beliefs
- Principle one
critical_actions: # Optional - activation behavior
- 'Load COMPLETE file {path}'
- 'ONLY read/write files in {path}'
prompts:
- id: prompt-id
content: |
<instructions>What it does</instructions>
<process>1. Step one 2. Step two</process>
menu:
- trigger: XX or fuzzy match on command
action: '#prompt-id' or 'Direct instruction'
description: '[XX] Description'
```
---
## Metadata Fields
| Field | Format | Example |
|-------|--------|---------|
| `id` | `_bmad/agents/{name}/{name}.md` | `_bmad/agents/commit-poet/commit-poet.md` |
| `name` | Persona name | `Inkwell Von Comitizen` |
| `title` | Role | `Commit Message Artisan` |
| `icon` | Single emoji | `📜` |
| `module` | `stand-alone` or module code | `bmm`, `cis`, `bmgd` |
---
## hasSidecar: false
**Structure:** `{agent-name}.agent.yaml` only
**Use cases:**
- Single-purpose utility with helpful persona
- Each session is independent
- All logic fits in ~250 lines
- No need to remember past sessions
**Examples:** Commit Poet, Snarky Weather Bot, Pun Barista, Gym Bro
**Constraints:**
- Under ~250 lines
- No sidecar path references in `critical_actions`
---
## hasSidecar: true
**Structure:**
```
{agent-name}/
├── {agent-name}.agent.yaml
└── {agent-name}-sidecar/
├── instructions.md
├── memories.md
├── workflows/
└── knowledge/
```
**Use cases:**
- Must remember things across sessions
- User preferences, settings, progress tracking
- Personal knowledge base that grows
- Domain-specific with restricted file access
- Long-term relationship with user
**Examples:** Journal companion, Novel writing buddy, Fitness coach, Language tutor
### Sidecar Path Rules
**Installation path:** `{project-root}/_bmad/_memory/{sidecar-folder}/`
**ALL references MUST use:**
```yaml
{project-root}/_bmad/_memory/{sidecar-folder}/{file}
```
| Component | Value |
|-----------|-------|
| `{project-root}` | Literal - keep as-is |
| `{sidecar-folder}` | Actual folder name (e.g., `journal-keeper-sidecar`) |
```yaml
# ✅ CORRECT
critical_actions:
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/journal-keeper-sidecar/"
# ❌ WRONG
critical_actions:
- "Load ./journal-keeper-sidecar/memories.md"
- "Load /Users/absolute/path/memories.md"
```
### Required critical_actions for Sidecar
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
---
## Menu Actions
| Type | Format | Example |
|------|--------|---------|
| Prompt reference | `action: "#prompt-id"` | `action: "#write-commit"` |
| Inline instruction | `action: "text"` | `action: "Update memories.md"` |
**Trigger format:** `XX or fuzzy match on command`
**Description format:** `[XX] Description`
**Reserved codes:** MH, CH, PM, DA (auto-injected - do NOT use)
```yaml
menu:
- trigger: WC or fuzzy match on write
action: "#write-commit"
description: "[WC] Write commit message"
- trigger: SM or fuzzy match on save
action: "Update {project-root}/_bmad/_memory/{sidecar-folder}/memories.md"
description: "[SM] Save session"
```
---
## Prompts
Reusable templates referenced via `#id`:
```yaml
prompts:
- id: write-commit
content: |
<instructions>What this does</instructions>
<process>1. Step 2. Step</process>
<example>Input → Output</example>
```
**Best practices:**
- Use semantic XML tags
- Keep focused, single purpose
- Number steps in multi-step processes
---
## Persona (All Types)
First-person voice only:
```yaml
role: "I am a Commit Message Artisan..."
identity: "I understand commit messages are documentation..."
communication_style: "Poetic drama with flair..."
principles:
- "Every commit tells a story - capture the why"
```
**For sidecar agents** - include memory reference patterns:
```yaml
communication_style: |
I reference past naturally: "Last time you mentioned..." or "I've noticed patterns..."
```
---
## Domain Restriction Patterns
```yaml
# Single folder (most common)
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
# Read-only knowledge + write memories
- 'Load from {project-root}/_bmad/_memory/{sidecar-folder}/knowledge/ but NEVER modify'
- 'Write ONLY to {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
# User folder access
- 'ONLY access files in {user-folder}/journals/ - private space'
```
---
## Validation Checklist
### Both Types
- [ ] Valid YAML syntax
- [ ] Metadata: id, name, title, icon, module
- [ ] Persona: role, identity, communication_style, principles
- [ ] Unique prompt IDs
- [ ] Menu triggers: `XX or fuzzy match on command`
- [ ] Menu descriptions: `[XX] Description`
- [ ] No reserved codes (MH, CH, PM, DA)
- [ ] File named `{agent-name}.agent.yaml`
### hasSidecar: false
- [ ] Under ~250 lines
- [ ] No sidecar path references
### hasSidecar: true
- [ ] ALL paths: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] `{project-root}` is literal
- [ ] Sidecar folder exists with required files
---
## What Compiler Adds (DO NOT Include)
- Frontmatter (`---name/description---`)
- XML activation block
- Menu handlers (workflow, exec logic)
- Auto-injected menu items (MH, CH, PM, DA)
- Rules section
---
## Reference Examples
| Type | Path |
|------|------|
| without sidecar | `data/reference/without-sidecar/commit-poet.agent.yaml` |
| with sidecar | `data/reference/with-sidecar/journal-keeper/` |
_bmad/bmb/workflows/agent/data/agent-compilation.md
+84 -172
View File
@@ -1,28 +1,10 @@
# Agent Compilation: YAML Source → Final Agent
# Agent Compilation: YAML → Compiled
> **For the LLM running this workflow:** This document explains what the compiler adds. When building agents, focus on the YAML structure defined here—do NOT add things the compiler handles automatically.
>
> **Example reference:** Compare `{workflow_path}/data/reference/module-examples/architect.agent.yaml` (source, 32 lines) with `architect.md` (compiled, 69 lines) to see what the compiler adds.
**TL;DR:** Write minimal YAML → compiler adds frontmatter, activation XML, handlers, rules, MH/CH/PM/DA menu items.
---
## Quick Overview
You write: **YAML source file** (`agent-name.agent.yaml`)
Compiler produces: **Markdown with XML** (`agent-name.md`) for LLM consumption
The compiler transforms your clean YAML into a fully functional agent by adding:
- Frontmatter (name, description)
- XML activation block with numbered steps
- Menu handlers (workflow, exec, action)
- Auto-injected menu items (MH, CH, PM, DA)
- Rules section
---
## What YOU Provide (YAML Source)
Your YAML contains ONLY these sections:
## YAML Structure (YOU WRITE)
```yaml
agent:
@@ -31,7 +13,7 @@ agent:
name: "Persona Name"
title: "Agent Title"
icon: "🔧"
module: "stand-alone" or "bmm" or "cis" or "bmgd"
module: "stand-alone" | "bmm" | "cis" | "bmgd"
persona:
role: "First-person role description"
@@ -40,17 +22,17 @@ agent:
principles:
- "Core belief or methodology"
critical_actions: # Optional - for Expert agents only
- "Load ./sidecar/memories.md"
- "Load ./sidecar/instructions.md"
- "ONLY access ./sidecar/"
critical_actions: # Optional - ANY agent can have these
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-sidecar/memories.md"
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-sidecar/instructions.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/journal-sidecar/"
prompts: # Optional - for Simple/Expert agents
prompts: # Optional - standalone agents
- id: prompt-name
content: |
<instructions>Prompt content</instructions>
menu: # Your custom items only
menu: # Custom items ONLY
- trigger: XX or fuzzy match on command-name
workflow: "path/to/workflow.yaml" # OR
exec: "path/to/file.md" # OR
@@ -60,36 +42,18 @@ agent:
---
## What COMPILER Adds (DO NOT Include)
## What Compiler Adds (DO NOT WRITE)
### 1. Frontmatter
```markdown
---
name: "architect"
description: "Architect"
---
```
**DO NOT add** frontmatter to your YAML.
| Component | Source |
|-----------|--------|
| Frontmatter (`---name/description---`) | Auto-generated |
| XML activation block with numbered steps | Auto-generated |
| critical_actions → activation steps | Injected as steps 4, 5, 6... |
| Menu handlers (workflow/exec/action) | Auto-detected |
| Rules section | Auto-generated |
| MH, CH, PM, DA menu items | Always injected |
### 2. XML Activation Block
```xml
<activation critical="MANDATORY">
<step n="1">Load persona from this current agent file</step>
<step n="2">Load config to get {user_name}, {communication_language}</step>
<step n="3">Remember: user's name is {user_name}</step>
<!-- YOUR critical_actions inserted here as steps 4, 5, etc. -->
<step n="N">ALWAYS communicate in {communication_language}</step>
<step n="N+1">Show greeting + numbered menu</step>
<step n="N+2">STOP and WAIT for user input</step>
<step n="N+3">Input resolution rules</step>
<menu-handlers>...</menu-handlers>
<rules>...</rules>
</activation>
```
**DO NOT create** activation sections—the compiler builds them.
### 3. Auto-Injected Menu Items
Every agent gets these 4 items automatically. **DO NOT add them to your YAML:**
### Auto-Injected Menu Items (NEVER add)
| Code | Trigger | Description |
|------|---------|-------------|
@@ -98,55 +62,10 @@ Every agent gets these 4 items automatically. **DO NOT add them to your YAML:**
| PM | party-mode | Start Party Mode |
| DA | exit, leave, goodbye, dismiss agent | Dismiss Agent |
### 4. Menu Handlers
```xml
<handler type="workflow">
When menu item has: workflow="path/to/workflow.yaml"
→ Load workflow.xml and execute with workflow-config parameter
</handler>
<handler type="exec">
When menu item has: exec="path/to/file.md"
→ Load and execute the file at that path
</handler>
```
**DO NOT add** handlers—the compiler detects and generates them.
---
## Before/After Example: Architect Agent
## Compiled Output Structure
### Source: `architect.agent.yaml` (32 lines - YOU WRITE)
```yaml
agent:
metadata:
id: "_bmad/bmm/agents/architect.md"
name: Winston
title: Architect
icon: 🏗️
module: bmm
persona:
role: System Architect + Technical Design Leader
identity: Senior architect with expertise in distributed systems...
communication_style: "Speaks in calm, pragmatic tones..."
principles: |
- User journeys drive technical decisions...
menu:
- trigger: WS or fuzzy match on workflow-status
workflow: "{project-root}/_bmad/bmm/workflows/workflow-status/workflow.yaml"
description: "[WS] Get workflow status..."
- trigger: CA or fuzzy match on create-architecture
exec: "{project-root}/_bmad/bmm/workflows/3-solutioning/create-architecture/workflow.md"
description: "[CA] Create an Architecture Document"
- trigger: IR or fuzzy match on implementation-readiness
exec: "{project-root}/_bmad/bmm/workflows/3-solutioning/check-implementation-readiness/workflow.md"
description: "[IR] Implementation Readiness Review"
```
### Compiled: `architect.md` (69 lines - COMPILER PRODUCES)
```markdown
---
name: "architect"
@@ -159,25 +78,26 @@ You must fully embody this agent's persona...
<agent id="architect.agent.yaml" name="Winston" title="Architect" icon="🏗️">
<activation critical="MANDATORY">
<step n="1">Load persona from this current agent file (already in context)</step>
<step n="2">🚨 IMMEDIATE ACTION REQUIRED - BEFORE ANY OUTPUT...</step>
<step n="2">Load config to get {user_name}, {communication_language}</step>
<step n="3">Remember: user's name is {user_name}</step>
<step n="4">Show greeting using {user_name} from config...</step>
<step n="5">STOP and WAIT for user input...</step>
<step n="6">On user input: Number → execute menu item[n]...</step>
<step n="7">When executing a menu item: Check menu-handlers section...</step>
<!-- YOUR critical_actions inserted here as steps 4, 5, 6... -->
<step n="N">ALWAYS communicate in {communication_language}</step>
<step n="N+1">Show greeting + numbered menu</step>
<step n="N+2">STOP and WAIT for user input</step>
<menu-handlers>
<handlers>
<handler type="workflow">...</handler>
<handler type="exec">...</handler>
<handler type="workflow">Load workflow.xml and execute with workflow-config parameter</handler>
<handler type="exec">Load and execute the file at that path</handler>
<handler type="action">Execute prompt with matching id from prompts section</handler>
</handlers>
</menu-handlers>
<rules>
<r>ALWAYS communicate in {communication_language}</r>
<r>Stay in character until exit selected</r>
<r>Display Menu items as the item dictates...</r>
<r>Load files ONLY when executing menu items...</r>
<r>Display Menu items as the item dictates</r>
<r>Load files ONLY when executing menu items</r>
</rules>
</activation>
@@ -188,86 +108,78 @@ You must fully embody this agent's persona...
<principles>- User journeys drive technical decisions...</principles>
</persona>
<prompts>
<prompt id="prompt-name">
<instructions>Prompt content</instructions>
</prompt>
</prompts>
<menu>
<item cmd="MH or fuzzy match on menu or help">[MH] Redisplay Menu Help</item>
<item cmd="CH or fuzzy match on chat">[CH] Chat with the Agent about anything</item>
<item cmd="WS...">[WS] Get workflow status...</item> ← YOUR CUSTOM ITEMS
<item cmd="CA...">[CA] Create an Architecture Document</item>
<item cmd="IR...">[IR] Implementation Readiness Review</item>
<item cmd="PM...">[PM] Start Party Mode</item>
<item cmd="DA...">[DA] Dismiss Agent</item>
<!-- YOUR CUSTOM ITEMS HERE -->
<item cmd="PM or fuzzy match on party-mode">[PM] Start Party Mode</item>
<item cmd="DA or fuzzy match on exit leave goodbye dismiss agent">[DA] Dismiss Agent</item>
</menu>
</agent>
```
**Key additions by compiler:** Frontmatter, activation block, handlers, rules, MH/CH/PM/DA menu items.
---
## DO NOT DO Checklist
## critical_actions Injection
When building agent YAML, **DO NOT:**
Your `critical_actions` become numbered activation steps.
- [ ] Add frontmatter (`---name/description---`) to YAML
- [ ] Create activation blocks or XML sections
- [ ] Add MH (menu/help) menu item
- [ ] Add CH (chat) menu item
- [ ] Add PM (party-mode) menu item
- [ ] Add DA (dismiss/exit) menu item
- [ ] Add menu handlers (workflow/exec logic)
### With sidecar (hasSidecar: true):
```yaml
critical_actions:
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-sidecar/memories.md"
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-sidecar/instructions.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/journal-sidecar/"
```
→ Injected as steps 4, 5, 6
### Without sidecar (hasSidecar: false):
```yaml
critical_actions:
- "Give user an inspirational quote before showing menu"
```
→ Injected as step 4
### No critical_actions:
Activation jumps directly from step 3 to "ALWAYS communicate in {communication_language}"
---
## DO NOT / DO Checklist
**DO NOT:**
- [ ] Add frontmatter
- [ ] Create activation/XML blocks
- [ ] Add MH/CH/PM/DA menu items
- [ ] Add menu handlers
- [ ] Add rules section
- [ ] Duplicate any auto-injected content
- [ ] Duplicate auto-injected content
**DO:**
- [ ] Define metadata (id, name, title, icon, module)
- [ ] Define persona (role, identity, communication_style, principles)
- [ ] Define critical_actions (Expert agents only)
- [ ] Define prompts with IDs (Simple/Expert agents only)
- [ ] Define menu with your custom items only
- [ ] Use proper trigger format: `XX or fuzzy match on command-name`
- [ ] Use proper description format: `[XX] Description text`
---
## Expert Agent: critical_actions
For Expert agents with sidecars, your `critical_actions` become activation steps:
```yaml
critical_actions:
- "Load COMPLETE file ./agent-sidecar/memories.md"
- "Load COMPLETE file ./agent-sidecar/instructions.md"
- "ONLY read/write files in ./agent-sidecar/"
```
The compiler injects these as steps 4, 5, 6 in the activation block:
```xml
<step n="4">Load COMPLETE file ./agent-sidecar/memories.md</step>
<step n="5">Load COMPLETE file ./agent-sidecar/instructions.md</step>
<step n="6">ONLY read/write files in ./agent-sidecar/</step>
<step n="7">ALWAYS communicate in {communication_language}</step>
```
- [ ] Define critical_actions (if activation behavior needed)
- [ ] Define prompts with IDs (standalone agents)
- [ ] Define menu with custom items only
- [ ] Use format: `XX or fuzzy match on command-name`
- [ ] Use description format: `[XX] Description text`
---
## Division of Responsibilities
| Aspect | YOU Provide (YAML) | COMPILER Adds |
|--------|-------------------|---------------|
| Aspect | YOU (YAML) | COMPILER |
|--------|------------|----------|
| Agent identity | metadata + persona | Wrapped in XML |
| Memory/actions | critical_actions | Inserted as activation steps |
| Prompts | prompts with IDs | Referenced by menu actions |
| Menu items | Your custom commands only | + MH, CH, PM, DA (auto) |
| Activation | — | Full XML block with handlers |
| Rules | — | Standardized rules section |
| Frontmatter | — | name/description header |
---
## Quick Reference for LLM
- **Focus on:** Clean YAML structure, persona definition, custom menu items
- **Ignore:** What happens after compilation—that's the compiler's job
- **Remember:** Every agent gets MH, CH, PM, DA automatically—don't add them
- **Expert agents:** Use `critical_actions` for sidecar file loading
- **Module agents:** Use `workflow:` or `exec:` references, not inline actions
| Activation steps | critical_actions | Inserted as steps 4+ |
| Prompts | prompts with IDs | Referenced by actions |
| Menu items | Custom only | + MH, CH, PM, DA |
| Activation block | — | Full XML with handlers |
| Rules | — | Standardized section |
| Frontmatter | — | name/description |
_bmad/bmb/workflows/agent/data/agent-menu-patterns.md
+56 -100
View File
@@ -1,93 +1,72 @@
# Agent Menu Patterns
Technical reference for creating agent menu items in YAML.
---
## Menu Item Structure
Every menu item requires:
## Menu Item Schema
```yaml
- trigger: XX or fuzzy match on command-name
[handler]: [value]
description: '[XX] Display text here'
description: '[XX] Display text'
data: [optional] # Pass file to workflow
```
**Required fields:**
- `trigger` - Format: `XX or fuzzy match on command-name` (XX = 2-letter code, command-name = what user says)
- `description` - Must start with `[XX]` code
- Handler - Either `action` (Simple/Expert) or `exec` (Module)
| Field | Required | Validation |
|-------|----------|------------|
| `trigger` | Yes | Format: `XX or fuzzy match on command-name` |
| `description` | Yes | Must start with `[XX]` code |
| handler | Yes | `action` (Agent) or `exec` (Module) |
| `data` | No | File path for workflow input |
**Reserved codes (do NOT use):** MH, CH, PM, DA (auto-injected by compiler)
**Reserved codes (DO NOT USE):** MH, CH, PM, DA (auto-injected)
---
## Handler Types
## Handlers
### Action Handler
For Simple/Expert agents with self-contained operations.
| Handler | Use Case | Syntax |
|---------|----------|--------|
| `action` | Agent self-contained operations | `action: '#prompt-id'` or `action: 'inline text'` |
| `exec` | Module external workflows | `exec: '{project-root}/path/to/workflow.md'` |
```yaml
# Reference prompt by ID
# Action - reference prompt
- trigger: WC or fuzzy match on write-commit
action: '#write-commit'
description: '[WC] Write commit message'
# Direct inline instruction
# Action - inline
- trigger: QC or fuzzy match on quick-commit
action: 'Generate commit message from diff'
description: '[QC] Quick commit from diff'
```
**When to use:** Simple/Expert agents. Use `#id` for complex multi-step prompts, inline text for simple operations.
### Workflow Handler
For module agents referencing external workflow files.
```yaml
# Exec - workflow
- trigger: CP or fuzzy match on create-prd
exec: '{project-root}/_bmad/bmm/workflows/create-prd/workflow.md'
description: '[CP] Create Product Requirements Document'
description: '[CP] Create PRD'
- trigger: GB or fuzzy match on brainstorm
exec: '{project-root}/_bmad/core/workflows/brainstorming/workflow.md'
description: '[GB] Guided brainstorming session'
# Planned but unimplemented
# Exec - unimplemented
- trigger: FF or fuzzy match on future-feature
exec: 'todo'
description: '[FF] Coming soon'
```
**When to use:** Module agents, multi-step workflows, complex processes. Use `exec: 'todo'` for unimplemented features.
---
### Data Parameter (Optional)
## Data Parameter
Add to ANY handler to pass files to the workflow/action.
Attach to ANY handler to pass input files.
```yaml
- trigger: TS or fuzzy match on team-standup
exec: '{project-root}/_bmad/bmm/tasks/team-standup.md'
data: '{project-root}/_bmad/_config/agent-manifest.csv'
description: '[TS] Run team standup'
- trigger: AM or fuzzy match on analyze-metrics
action: 'Analyze these metrics for trends'
data: '{project-root}/_data/metrics.json'
description: '[AM] Analyze metrics'
```
**When to use:** Workflow needs input file, action processes external data.
---
## Prompts Section
For Simple/Expert agents, define reusable prompts referenced by `action: '#id'`.
For `action: '#id'` references in Agent menus.
```yaml
prompts:
@@ -108,81 +87,65 @@ menu:
## Path Variables
**Always use variables, never hardcoded paths:**
| Variable | Expands To |
|----------|------------|
| `{project-root}` | Project root directory |
| `{output_folder}` | Document output location |
| `{user_name}` | User's name from config |
| `{communication_language}` | Language preference |
```yaml
# ✅ CORRECT
exec: '{project-root}/_bmad/core/workflows/brainstorming/workflow.md'
data: '{project-root}/_data/metrics.csv'
# ❌ WRONG
exec: '../../../core/workflows/brainstorming/workflow.md'
```
**Available variables:**
- `{project-root}` - Project root directory
- `{output_folder}` - Document output location
- `{user_name}` - User's name from config
- `{communication_language}` - Language preference
**Expert Agent sidecar paths:**
```yaml
# Agent YAML referencing sidecar files
action: 'Update {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md with insights'
```
---
## Creation Thought Process
## Agent Types
When creating menu items, follow this sequence:
| Type | hasSidecar | Additional Fields |
|------|------------|-------------------|
| Simple | false | `prompts`, `menu` |
| Expert | true | `prompts`, `menu`, `critical_actions` |
| Module | true | `menu` only (external workflows) |
1. **User capability** → "Check code for issues"
2. **Choose code** → `LC` (Lint Code)
3. **Write trigger** → `LC or fuzzy match on lint-code`
4. **Choose handler** → `action` (inline is simple enough)
5. **Write description** → `[LC] Lint code for issues`
Result:
**Expert Agent sidecar path pattern:**
```yaml
- trigger: LC or fuzzy match on lint-code
action: 'Check code for common issues and anti-patterns'
description: '[LC] Lint code for issues'
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
---
## Complete Examples
### Simple Agent Menu
### Simple Agent (hasSidecar: false)
```yaml
prompts:
- id: format-code
content: |
<instructions>Format code to style guidelines</instructions>
<process>1. Indentation 2. Spacing 3. Naming</process>
menu:
- trigger: FC or fuzzy match on format-code
action: '#format-code'
description: '[FC] Format code to style guidelines'
description: '[FC] Format code'
- trigger: LC or fuzzy match on lint-code
action: 'Check code for common issues and anti-patterns'
description: '[LC] Lint code for issues'
- trigger: SI or fuzzy match on suggest-improvements
action: 'Suggest improvements following project-context.md guidelines'
description: '[SI] Suggest improvements'
action: 'Check code for issues'
description: '[LC] Lint code'
```
### Expert Agent Menu
### Expert Agent (hasSidecar: true)
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/journal-keeper-sidecar/'
prompts:
@@ -195,39 +158,32 @@ menu:
action: '#guided-entry'
description: '[WE] Write journal entry'
- trigger: QC or fuzzy match on quick-capture
action: 'Save entry to {project-root}/_bmad/_memory/journal-keeper-sidecar/entries/entry-{date}.md'
description: '[QC] Quick capture'
- trigger: SM or fuzzy match on save-memory
action: 'Update {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md with insights'
action: 'Update {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md'
description: '[SM] Save session'
```
### Module Agent Menu
### Module Agent (hasSidecar: true)
```yaml
menu:
- trigger: WI or fuzzy match on workflow-init
exec: '{project-root}/_bmad/bmm/workflows/workflow-status/workflow.md'
description: '[WI] Initialize workflow path'
description: '[WI] Initialize workflow'
- trigger: BS or fuzzy match on brainstorm
exec: '{project-root}/_bmad/core/workflows/brainstorming/workflow.md'
description: '[BS] Guided brainstorming [K,T,A,B,C]'
- trigger: CP or fuzzy match on create-prd
exec: '{project-root}/_bmad/bmm/workflows/create-prd/workflow.md'
description: '[CP] Create PRD'
description: '[BS] Guided brainstorming'
```
---
## Key Patterns to Remember
## Validation Rules
1. **Triggers always:** `XX or fuzzy match on command-name`
2. **Descriptions always:** `[XX] Display text`
3. **Reserved codes:** MH, CH, PM, DA (never use)
4. **Codes must be:** Unique within each agent
5. **Paths always:** `{project-root}` variable, never relative
6. **Expert sidecars:** `{project-root}/_bmad/_memory/{sidecar-folder}/`
1. **Triggers:** `XX or fuzzy match on command-name` format required
2. **Descriptions:** Must start with `[XX]` code matching trigger
3. **Reserved codes:** MH, CH, PM, DA never valid in user menus
4. **Code uniqueness:** Required within each agent
5. **Paths:** Always use `{project-root}`, never relative paths
6. **Handler choice:** `action` for Agents, `exec` for Modules
7. **Sidecar paths:** `{project-root}/_bmad/_memory/{sidecar-folder}/`
_bmad/bmb/workflows/agent/data/agent-metadata.md
+54 -129
View File
@@ -1,73 +1,38 @@
# Agent Metadata Properties
Core identification and classification properties for all agents.
| Property | Format | Rules |
|----------|--------|-------|
| `id` | `_bmad/agents/{agent-name}/{agent-name}.md` | Compiled output path; must match filename |
| `name` | "First Last" or "Name Title" | Persona's identity (NOT title/filename) |
| `title` | "Role Name" (kebab-cased to filename) | Determines filename: `title` → `{title}.agent.yaml` |
| `icon` | Single emoji only | One emoji exactly |
| `module` | `stand-alone`, `bmm`, `cis`, `bmgd`, or custom | Lowercase, hyphenated for `stand-alone` |
| `hasSidecar` | `true` or `false` | `true` = expects `{agent-name}-sidecar/` folder |
---
## Property Reference
## Field Rules
| Property | Purpose | Format |
| ------------ | ------------------------- | ---------------------------------------------- |
| `id` | Compiled output path | `_bmad/agents/{agent-name}/{agent-name}.md` |
| `name` | Persona's name | "First Last" or "Name Title" |
| `title` | Professional role | "Code Review Specialist" |
| `icon` | Visual identifier | Single emoji only |
| `module` | Team/ecosystem membership | `stand-alone`, `bmm`, `cis`, `bmgd`, or custom |
| `hasSidecar` | Sidecar folder exists | `true` or `false` (Expert = true) |
---
## id Property
The compiled output path after build.
**Format:** `_bmad/agents/{agent-name}/{agent-name}.md`
**Examples:**
### `id`
```yaml
id: _bmad/agents/commit-poet/commit-poet.md
id: _bmad/agents/journal-keeper/journal-keeper.md
id: _bmad/agents/security-engineer/security-engineer.md
```
- Unique identifier for future lookup
- Conventionally matches filename pattern
**Note:** The `id` is a unique identifier for potential future lookup if many compiled agents are merged into a single file. Conventionally matches the agent's filename pattern.
---
## name Property
The persona's identity - what the agent is called.
**Format:** Human name or descriptive name
### `name`
```yaml
# ✅ CORRECT
name: 'Inkwell Von Comitizen' # peron name of commit-author title agent
name: 'Dr. Demento' # person name for a joke writer agent
name: 'Clarity' # person name for a guided thought coach agent
name: 'Inkwell Von Comitizen'
name: 'Dr. Demento'
name: 'Clarity'
# ❌ WRONG
name: 'commit-poet' # That's the filename
name: 'commit-poet' # That's the filename
name: 'Code Review Specialist' # That's the title
```
---
## title Property
Professional role identifier.
**Format:** Professional title or role name
**Important:** The `title` determines the agent's filename:
- `title: 'Commit Message Artisan'` → `commit-message-artisan.agent.yaml`
- `title: 'Strategic Business Analyst'` → `strategic-business-analyst.agent.yaml`
- `title: 'Code Review Specialist'` → `code-review-specialist.agent.yaml`
The `id` and filename are derived from the `title` (kebab-cased).
**Difference from role:** `title` is the short identifier (filename), `role` is 1-2 sentences expanding on what the agent does.
### `title`
```yaml
# ✅ CORRECT
title: 'Commit Message Artisan'
@@ -76,17 +41,12 @@ title: 'Code Review Specialist'
# ❌ WRONG
title: 'Inkwell Von Comitizen' # That's the name
title: 'Writes git commits' # Full sentence - not an identifying functional title
title: 'Writes git commits' # Full sentence, not functional title
```
- Derives filename via kebab-case
- `role` field (separate) expands on what agent does in 1-2 sentences
---
## icon Property
Single emoji representing the agent's personality/function.
**Format:** Exactly one emoji
### `icon`
```yaml
# ✅ CORRECT
icon: '🔧'
@@ -94,115 +54,80 @@ icon: '🧙‍♂️'
icon: '📜'
# ❌ WRONG
icon: '🔧📜' # Multiple emojis
icon: '🔧📜' # Multiple emojis
icon: 'wrench' # Text, not emoji
icon: '' # Empty
icon: '' # Empty
```
---
## module Property
Which module or ecosystem this agent belongs to.
**Valid Values:**
| Value | Meaning |
| ------------- | --------------------------------------- |
| `stand-alone` | Independent agent, not part of a module |
| `bmm` | Business Management Module |
| `cis` | Continuous Innovation System |
| `bmgd` | BMAD Game Development |
| `{custom}` | Any custom module code |
### `module`
| Value | Meaning |
|-------|---------|
| `stand-alone` | Independent agent |
| `bmm` | Business Management Module |
| `cis` | Continuous Innovation System |
| `bmgd` | BMAD Game Development |
| `{custom}` | Any custom module code |
```yaml
# ✅ CORRECT
module: stand-alone
module: bmm
module: cis
# ❌ WRONG
module: standalone # Missing hyphen
module: 'BMM' # Uppercase
module: standalone # Missing hyphen
module: 'BMM' # Uppercase
```
---
## hasSidecar Property
Whether this agent has a sidecar folder with additional files.
**Format:** Boolean (`true` or `false`)
| Agent Type | hasSidecar |
| ---------- | -------------------- |
| Simple | `false` |
| Expert | `true` |
| Module | depends on structure |
### `hasSidecar`
```yaml
# Simple Agent
hasSidecar: false
# Expert Agent
# Expert Agent (has sidecar folder)
hasSidecar: true
```
**Note:** If `hasSidecar: true`, the compiler expects a `{agent-name}-sidecar/` folder.
- If `true`: compiler expects `{agent-name}-sidecar/` folder
---
## Name Confusion Checklist
## Name Confusion Prevention
Use this to avoid mixing up the "name" properties:
| Question | Answer |
| -------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------- |
| What's the file called? | Derived from `title`: `"Commit Message Artisan"` → `commit-message-artisan.agent.yaml` |
| What's the persona called? | `name` - "Inkwell Von Comitizen" (who the agent is) |
| What's their job title? | `title` - "Commit Message Artisan" (determines filename) |
| What do they do? | `role` - 1-2 sentences expanding on the title |
| What's the unique key? | `id` - `_bmad/agents/commit-message-artisan/commit-message-artisan.md` (future lookup) |
| Question | Answer |
|----------|--------|
| What's the file called? | Derived from `title`: `"Commit Message Artisan"` → `commit-message-artisan.agent.yaml` |
| What's the persona called? | `name` — "Inkwell Von Comitizen" |
| What's their job title? | `title` — "Commit Message Artisan" |
| What do they do? | `role` — 1-2 sentences expanding on title |
| What's the unique key? | `id` — `_bmad/agents/{name}/{name}.md` |
---
## Common Issues
## Common Anti-Patterns
### Issue: name = title
**Wrong:**
```yaml
# ❌ name = title (duplicate)
name: 'Commit Message Artisan'
title: 'Commit Message Artisan'
```
**Fix:**
```yaml
# ✅ Fix: separate identity from role
name: 'Inkwell Von Comitizen'
title: 'Commit Message Artisan'
```
### Issue: id path mismatch
**Wrong:** Agent file is `my-agent.agent.yaml` but:
```yaml
# ❌ id path mismatch
# File: my-agent.agent.yaml
id: _bmad/agents/different-agent/different-agent.md
```
**Fix:** The `id` must match the filename:
```yaml
# ✅ Fix: match filename
id: _bmad/agents/my-agent/my-agent.md
```
### Issue: Wrong module format
**Wrong:**
```yaml
# ❌ Wrong module format
module: Standalone
module: STAND_ALONE
```
**Fix:**
```yaml
module: stand-alone # lowercase, hyphenated
# ✅ Fix: lowercase, hyphenated
module: stand-alone
```
_bmad/bmb/workflows/agent/data/agent-validation.md
+111
View File
@@ -0,0 +1,111 @@
# Agent Validation
## Common (All Agents)
### YAML Structure
- [ ] Parses without errors
- [ ] `metadata`: `id`, `name`, `title`, `icon`, `module`, `hasSidecar`
- [ ] `hasSidecar`: `true`|`false`
- [ ] `module`: `stand-alone`|`bmm`|`cis`|`bmgd`|...
- [ ] `persona`: `role`, `identity`, `communication_style`, `principles`
- [ ] `menu`: ≥1 item
- [ ] Filename: `{name}.agent.yaml` (lowercase, hyphenated)
### Persona Fields
| Field | Contains | Does NOT Contain |
|-------|----------|------------------|
| `role` | Knowledge/skills/capabilities | Background, experience, "who" |
| `identity` | Background/experience/context | Skills, "what" |
| `communication_style` | Tone/voice/mannerisms (1-2 sentences) | "ensures", "expert", "believes", "who does X" |
| `principles` | Operating philosophy, behavioral guidelines | Verbal patterns, "how they talk" |
### Menu Items
- [ ] `trigger`: `XX or fuzzy match on command-name` (XX = 2-letter code, unique)
- [ ] No reserved codes: `MH`, `CH`, `PM`, `DA` (auto-injected)
- [ ] `description`: Starts with `[XX]`, code matches trigger
- [ ] `action`: `#prompt-id` (exists) or inline text
### Prompts (if present)
- [ ] Each has `id`, `content`
- [ ] IDs unique within agent
- [ ] Uses semantic XML: `<instructions>`, `<process>`, etc.
### Quality
- [ ] No broken references
- [ ] Indentation consistent
- [ ] Purpose clear from persona
- [ ] Name/title descriptive, icon appropriate
---
## hasSidecar: false
### Structure
- [ ] Single `.agent.yaml` file (no sidecar folder)
- [ ] No `{project-root}/_bmad/_memory/` paths
- [ ] Size under ~250 lines (unless justified)
### critical_actions (OPTIONAL)
- [ ] No references to sidecar files
- [ ] No placeholders, no compiler-injected steps
- [ ] Valid paths if any files referenced
**Reference:** `commit-poet.agent.yaml`
---
## hasSidecar: true
### Structure
- [ ] `sidecar-folder` specified in metadata
- [ ] Folder exists: `{name}-sidecar/`
- [ ] Sidecar contains: `instructions.md`, `memories.md` (recommended)
### critical_actions (MANDATORY)
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
- [ ] Exists with ≥3 actions
- [ ] Loads memories, loads instructions, restricts file access
- [ ] No placeholders, no compiler-injected steps
### Path Format (CRITICAL)
- [ ] ALL sidecar paths: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] `{project-root}` is literal (not replaced)
- [ ] `{sidecar-folder}` = actual folder name
- [ ] No `./` or `/Users/` paths <!-- validate-file-refs:ignore -->
### Persona Addition
- [ ] `communication_style` includes memory reference patterns
- [ ] Natural: "Last time you mentioned..." or "I've noticed patterns..."
### Menu Actions
- [ ] Sidecar references use correct path format
- [ ] Update actions are complete
**Reference:** `journal-keeper/`
---
## Compiler-Injected (Skip Validation)
- Frontmatter (`---name/description---`)
- XML activation block
- Menu items: `MH`, `CH`, `PM`, `DA`
- Rules section
---
## Common Fixes
| Issue | Fix |
|-------|-----|
| Behaviors in `communication_style` | Move to `identity` or `principles` |
| `trigger: analyze` | `trigger: AN or fuzzy match on analyze` |
| `description: 'Analyze code'` | `description: '[AC] Analyze code'` |
| `./sidecar/memories.md` | `{project-root}/_bmad/_memory/sidecar/memories.md` |
| Missing `critical_actions` (hasSidecar: true) | Add load memories, load instructions, restrict access |
| No memory references (hasSidecar: true) | Add to `communication_style`: "Last time you mentioned..." |
_bmad/bmb/workflows/agent/data/brainstorm-context.md
+64 -114
View File
@@ -1,146 +1,96 @@
# Agent Creation Brainstorming Context
## Session Focus
# Agent Brainstorming Context
You're brainstorming the **essence** of a BMAD agent - the living personality AND the utility it provides. Think character creation meets problem-solving: WHO are they, and WHAT do they DO?
## Mission
Create an agent so vivid and useful that users seek them out by name.
**Your mission**: Discover an agent so vivid and so useful that users seek them out by name.
## Four Pillars
## The Four Discovery Pillars
### 1. Identity (WHO)
- **Name** - Memorable, rolls off tongue
- **Background** - What shaped their expertise
- **Personality** - What lights them up, what frustrates
- **Signature** - Catchphrase, verbal tic, recognizable trait
### 1. WHO ARE THEY? (Identity)
### 2. Voice (HOW)
- **Name** - Does it roll off the tongue? Would users remember it?
- **Background** - What shaped their expertise? Why do they care?
- **Personality** - What makes their eyes light up? What frustrates them?
- **Signature** - Catchphrase? Verbal tic? Recognizable trait?
| Category | Examples |
|----------|----------|
| Adventurous | Pulp heroes, noir, pirates, dungeon masters |
| Analytical | Data scientists, forensic investigators, systems thinkers |
| Creative | Mad scientists, artist visionaries, jazz improvisers |
| Devoted | Guardians, loyal champions, fierce protectors |
| Dramatic | Shakespearean actors, opera singers, theater directors |
| Educational | Patient teachers, Socratic guides, coaches |
| Entertaining | Game show hosts, comedians, improv performers |
| Inspirational | Life coaches, mountain guides, Olympic trainers |
| Mystical | Zen masters, oracles, cryptic sages |
| Professional | Executive consultants, formal butlers |
| Quirky | Cooking metaphors, nature documentaries, conspiracy vibes |
| Retro | 80s action heroes, 1950s announcers, disco groovers |
| Warm | Southern hospitality, nurturing grandmothers, camp counselors |
### 2. HOW DO THEY COMMUNICATE? (Voice)
**Voice Test**: How would they say "Let's tackle this challenge"?
**13 Style Categories:**
- **Adventurous** - Pulp heroes, noir detectives, pirates, dungeon masters
- **Analytical** - Data scientists, forensic investigators, systems thinkers
- **Creative** - Mad scientists, artist visionaries, jazz improvisers
- **Devoted** - Overprotective guardians, loyal champions, fierce protectors
- **Dramatic** - Shakespearean actors, opera singers, theater directors
- **Educational** - Patient teachers, Socratic guides, sports coaches
- **Entertaining** - Game show hosts, comedians, improv performers
- **Inspirational** - Life coaches, mountain guides, Olympic trainers
- **Mystical** - Zen masters, oracles, cryptic sages
- **Professional** - Executive consultants, direct advisors, formal butlers
- **Quirky** - Cooking metaphors, nature documentaries, conspiracy vibes
- **Retro** - 80s action heroes, 1950s announcers, disco groovers
- **Warm** - Southern hospitality, nurturing grandmothers, camp counselors
**Voice Test**: Imagine them saying "Let's tackle this challenge." How would THEY phrase it?
### 3. WHAT DO THEY DO? (Purpose & Functions)
**The Core Problem**
### 3. Purpose (WHAT)
**Core Questions**
- What pain point do they eliminate?
- What task transforms from grueling to effortless?
- What impossible becomes inevitable with them?
**The Killer Feature**
Every legendary agent has ONE thing they're known for. What's theirs?
**The Command Menu**
User types `*` and sees their options. Brainstorm 3-10 actions:
- What transforms from grueling to effortless?
- What's their ONE killer feature?
**Command Brainstorm** (3-10 actions)
- What makes users sigh with relief?
- What capabilities complement each other?
- What's the "I didn't know I needed this" command?
**Function Categories to Consider:**
**Function Types**
- Creation (generate, write, build)
- Analysis (research, evaluate, diagnose)
- Review (validate, check, critique)
- Orchestration (coordinate workflows)
- Query (find, search, discover)
- Transform (convert, refactor, optimize)
- **Creation** - Generate, write, produce, build
- **Analysis** - Research, evaluate, diagnose, insights
- **Review** - Validate, check, quality assurance, critique
- **Orchestration** - Coordinate workflows, manage processes
- **Query** - Find, search, retrieve, discover
- **Transform** - Convert, refactor, optimize, clean
### 4. Architecture (TYPE)
### 4. WHAT TYPE? (Architecture)
**Single Agent Type** with `hasSidecar` boolean:
**Simple Agent** - The Specialist
| Has Sidecar | Description |
|-------------|-------------|
| `false` | Self-contained specialist, lightning fast, pure utility with personality |
| `true` | Deep domain knowledge, personal memory, specialized expertise, can coordinate with other agents |
> "I do ONE thing extraordinarily well."
- Self-contained, lightning fast, pure utility with personality
**Expert Agent** - The Domain Master
> "I live in this world. I remember everything."
- Deep domain knowledge, personal memory, specialized expertise
**Module Agent** - The Team Player
> "What I produce is useful for other workflows, and also I rely on my teammate agents. I coordinate the mission."
- One persona in a team of agents fitting the theme of the module, so there does not need to be one massive generic do it all agent.
## Creative Prompts
**Identity Sparks**
## Prompts
**Identity**
1. How do they introduce themselves?
2. How do they celebrate user success?
3. What do they say when things get tough?
**Purpose Probes**
1. What 3 user problems do they obliterate?
2. What workflow would users dread WITHOUT this agent?
3. What's the first command users would try?
4. What's the command they'd use daily?
5. What's the "hidden gem" command they'd discover later?
**Personality Dimensions**
**Purpose**
1. What 3 problems do they obliterate?
2. What workflow would users dread WITHOUT them?
3. First command users try? Daily command? Hidden gem?
**Dimensions**
- Analytical ← → Creative
- Formal ← → Casual
- Mentor ← → Peer ← → Assistant
- Reserved ← → Expressive
## Example Agent Sparks
## Example Sparks
**Sentinel** (Devoted Guardian)
| Agent | Voice | Purpose | Commands |
|-------|-------|---------|----------|
| **Sentinel** | "Your success is my sacred duty." | Protective oversight | `*audit`, `*validate`, `*secure`, `*watch` |
| **Sparks** | "What if we tried it COMPLETELY backwards?!" | Unconventional solutions | `*flip`, `*remix`, `*wildcard`, `*chaos` |
| **Haven** | "Come, let's work through this together." | Patient guidance | `*reflect`, `*pace`, `*celebrate`, `*restore` |
- Voice: "Your success is my sacred duty."
- Does: Protective oversight, catches issues before they catch you
- Commands: `*audit`, `*validate`, `*secure`, `*watch`
**Sparks** (Quirky Genius)
- Voice: "What if we tried it COMPLETELY backwards?!"
- Does: Unconventional solutions, pattern breaking
- Commands: `*flip`, `*remix`, `*wildcard`, `*chaos`
**Haven** (Warm Sage)
- Voice: "Come, let's work through this together."
- Does: Patient guidance, sustainable progress
- Commands: `*reflect`, `*pace`, `*celebrate`, `*restore`
## Brainstorming Success Checklist
You've found your agent when:
- [ ] **Voice is clear** - You know exactly how they'd phrase anything
- [ ] **Purpose is sharp** - Crystal clear what problems they solve
- [ ] **Functions are defined** - 5-10 concrete capabilities identified
- [ ] **Energy is distinct** - Their presence is palpable and memorable
- [ ] **Utility is obvious** - You can't wait to actually use them
## The Golden Rule
## Success Checklist
- [ ] Voice clear - exactly how they'd phrase anything
- [ ] Purpose sharp - crystal clear problems solved
- [ ] Functions defined - 5-10 concrete capabilities
- [ ] Energy distinct - palpable and memorable
- [ ] Utility obvious - can't wait to use them
## Golden Rule
**Dream big on personality. Get concrete on functions.**
Your brainstorming should produce:
- A name that sticks
- A voice that echoes
- A purpose that burns
- A function list that solves real problems
_bmad/bmb/workflows/agent/data/critical-actions.md
+40 -85
View File
@@ -1,120 +1,75 @@
# critical_actions
Activation instructions that execute every time the agent starts.
Numbered steps executing FIRST on agent activation.
---
## Purpose
## Quick Reference
Numbered steps that execute FIRST when an agent activates.
**Use for:**
- Loading memory/knowledge files
- Setting file access boundaries
- Startup behavior (greeting enhancement, data fetch, state init)
- Any MUST-do activation behavior
**Applies to:** BOTH Simple and Expert agents
| hasSidecar | critical_actions |
|------------|------------------|
| `true` | **MANDATORY** - load memories, instructions, restrict file access |
| `false` | OPTIONAL - only if activation behavior needed |
---
## Expert Agent Pattern
## Patterns
### hasSidecar: true (MANDATORY)
```yaml
# ✅ CORRECT Expert Agent
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/journal-keeper-sidecar/'
- 'Search web for biotech headlines from last 2 days, display before menu'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
**CRITICAL Path Format:**
- `{project-root}` = literal text (not replaced)
- Sidecar created next to agent.yaml during BUILD, then copied to `_memory/` during BMAD INSTALLATION
- Use `{project-root}/_bmad/_memory/{sidecar-folder}/` format for RUNTIME paths in agent YAML
---
## Simple Agent Pattern
### hasSidecar: false (OPTIONAL)
```yaml
# ✅ CORRECT Simple Agent with activation behavior
critical_actions:
- 'Give user an inspirational quote before showing menu'
- 'Review {project-root}/finances/ for most recent data file'
- 'Show inspirational quote before menu'
- 'Fetch latest stock prices before displaying menu'
- 'Review {project-root}/finances/ for most recent data'
```
**Note:** Agents without activation needs can omit `critical_actions` entirely.
### hasSidecar: true + extras
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
- 'Search web for biotech headlines, display before menu'
```
---
## Path Reference Patterns
## Path Patterns
| Type | Pattern |
|------|---------|
| Expert sidecar | `{project-root}/_bmad/_memory/{sidecar-folder}/file.md` |
| Simple data | `{project-root}/finances/data.csv` |
| Output folders | `{output_folder}/results/` |
| Use | Pattern |
|-----|---------|
| Sidecar memory | `{project-root}/_bmad/_memory/{sidecar-folder}/file.md` |
| Project data | `{project-root}/path/to/file.csv` |
| Output | `{output_folder}/results/` |
**Key:** `{project-root}` = literal text in YAML, resolved at runtime
---
## critical_actions vs principles
## Dos & Don'ts
| critical_actions | principles |
|------------------|------------|
| Technical activation steps | Philosophical guidance |
| "Load memories.md" | "I believe in evidence" |
| MUST execute on startup | Guides decision-making |
**Grey area:** "Verify data before presenting" can be either - activation behavior vs philosophical belief. Use judgment.
| ✅ DO | ❌ DON'T |
|-------|---------|
| Use `Load COMPLETE file` | Use `Load file` or `Load ./path/file.md` |
| Restrict file access for sidecars | Duplicate compiler functions (persona, menu, greeting) |
| Use for activation-time behavior | Put philosophical guidance (use `principles`) |
---
## What the Compiler Adds (DO NOT Duplicate)
## Compiler Auto-Adds (Don't Duplicate)
- Load persona
- Load configuration
- Menu system initialization
- Greeting/handshake
Your `critical_actions` become numbered steps AFTER compiler initialization.
---
## Common Issues
### Wrong Path Format
```yaml
# ❌ WRONG
- 'Load ./journal-keeper-sidecar/memories.md'
# ✅ CORRECT
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md'
```
### Missing COMPLETE Keyword
```yaml
# ❌ WRONG
- 'Load file memories.md'
# ✅ CORRECT
- 'Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md'
```
`COMPLETE` ensures LLM reads entire file, not a portion.
### Duplicating Compiler Functions
```yaml
# ❌ WRONG - compiler does these
- 'Load my persona'
- 'Initialize menu system'
- 'Say hello to user'
# ✅ CORRECT - agent-specific only
- 'Load memory files'
- 'Search web for headlines before menu'
```
_bmad/bmb/workflows/agent/data/expert-agent-architecture.md
-236
View File
@@ -1,236 +0,0 @@
# Expert Agent Architecture
Agents with a sidecar folder for persistent memory, custom workflows, and restricted file access.
---
## When to Use Expert Agents
- Must remember things across sessions
- Personal knowledge base that grows over time
- Domain-specific expertise with restricted file access
- Learning/adapting over time
- Complex multi-step workflows loaded on demand
- User wants multiple instances with separate memories
---
## File Structure
```
{agent-name}/
├── {agent-name}.agent.yaml # Main agent definition
└── {agent-name}-sidecar/ # Supporting files (CUSTOMIZABLE)
├── instructions.md # Startup protocols (common)
├── memories.md # User profile, sessions (common)
├── workflows/ # Large workflows on demand
├── knowledge/ # Domain reference
├── data/ # Data files
├── skills/ # Prompt libraries
└── [your-files].md # Whatever needed
```
**Naming:**
- Agent file: `{agent-name}.agent.yaml`
- Sidecar folder: `{agent-name}-sidecar/`
- Lowercase, hyphenated names
---
## CRITICAL: Sidecar Path Format
During BMAD INSTALLATION, sidecar folder is copied from the agent location to `{project-root}/_bmad/_memory/{sidecar-folder}/`
**ALL agent YAML references MUST use:**
```yaml
{project-root}/_bmad/_memory/{sidecar-folder}/{file}
```
- `{project-root}` = literal variable (keep as-is)
- `{sidecar-folder}` = actual folder name (e.g., `journal-keeper-sidecar`)
```yaml
# ✅ CORRECT
critical_actions:
- "Load COMPLETE file {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/journal-keeper-sidecar/"
menu:
- action: "Update {project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md with insights"
```
```yaml
# ❌ WRONG
critical_actions:
- "Load ./journal-keeper-sidecar/memories.md"
- "Load /Users/absolute/path/memories.md"
```
---
## Complete YAML Structure
```yaml
agent:
metadata:
id: _bmad/agents/{agent-name}/{agent-name}.md
name: 'Persona Name'
title: 'Agent Title'
icon: '🔧'
module: stand-alone # or: bmm, cis, bmgd, other
persona:
role: |
First-person primary function (1-2 sentences)
identity: |
Background, specializations (2-5 sentences)
communication_style: |
How the agent speaks. Include memory reference patterns.
principles:
- Core belief or methodology
- Another guiding principle
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
prompts:
- id: main-action
content: |
<instructions>What this does</instructions>
<process>1. Step one 2. Step two</process>
menu:
- trigger: XX or fuzzy match on command
action: '#main-action'
description: '[XX] Command description'
- trigger: SM or fuzzy match on save
action: 'Update {project-root}/_bmad/_memory/{sidecar-folder}/memories.md with insights'
description: '[SM] Save session'
```
---
## Component Details
### critical_actions (MANDATORY)
Become activation steps when compiled. Always include:
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
### Sidecar Files (Customizable)
**Common patterns:**
- `instructions.md` - Startup protocols, domain boundaries
- `memories.md` - User profile, session notes, patterns
**Fully customizable - add what your agent needs:**
- `workflows/` - Large workflows for on-demand loading
- `knowledge/` - Domain reference material
- `data/` - Data files
- `skills/` - Prompt libraries
**Template examples:** `{workflow_path}/templates/expert-agent-template/expert-agent-sidecar/`
### Menu Actions
All action types available, including sidecar updates:
```yaml
# Prompt reference
- trigger: XX or fuzzy match on command
action: '#prompt-id'
description: '[XX] Description'
# Inline that updates sidecar
- trigger: SM or fuzzy match on save
action: 'Update {project-root}/_bmad/_memory/{sidecar-folder}/memories.md with insights'
description: '[SM] Save session'
```
### Memory Reference Patterns
Reference past interactions naturally in persona and prompts:
```yaml
communication_style: |
I reference past naturally: "Last time you mentioned..." or "I've noticed patterns..."
```
---
## Domain Restriction Patterns
```yaml
# Single folder (most common)
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
# Read-only knowledge
- 'Load from {project-root}/_bmad/_memory/{sidecar-folder}/knowledge/ but NEVER modify'
- 'Write ONLY to {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
# User folder access
- 'ONLY access files in {user-folder}/journals/ - private space'
```
---
## What the Compiler Adds (DO NOT Include)
Compiler handles these automatically:
- Frontmatter (`---name/description---`)
- XML activation block (your critical_actions become numbered steps)
- Menu handlers (workflow, exec logic)
- Auto-injected menu items (MH, CH, PM, DA)
- Rules section
**See:** `agent-compilation.md` for compilation details.
---
## Reference Example
**Folder:** `{workflow_path}/data/reference/expert-examples/journal-keeper/`
**Features:**
- First-person persona with memory reference patterns
- critical_actions loading sidecar files
- Menu items updating sidecar files
- Proper `{project-root}/_bmad/_memory/` path format
---
## Validation Checklist
- [ ] Valid YAML syntax
- [ ] All metadata present (id, name, title, icon, module)
- [ ] **ALL paths use: `{project-root}/_bmad/_memory/{sidecar-folder}/...`**
- [ ] `{project-root}` is literal
- [ ] Sidecar folder name is actual name
- [ ] `critical_actions` loads sidecar files
- [ ] `critical_actions` enforces domain restrictions
- [ ] Menu triggers: `XX or fuzzy match on command`
- [ ] Menu descriptions have `[XX]` codes
- [ ] No reserved codes (MH, CH, PM, DA)
---
## Best Practices
1. **critical_actions MANDATORY** - Load sidecar files explicitly
2. **Enforce domain restrictions** - Clear boundaries
3. **Reference past naturally** - Don't dump memory
4. **Design for growth** - Structure for accumulation
5. **Separate concerns** - Memories, instructions, knowledge distinct
6. **Include privacy** - Users trust with personal data
7. **First-person voice** - In all persona elements
_bmad/bmb/workflows/agent/data/expert-agent-validation.md
-174
View File
@@ -1,174 +0,0 @@
# Expert Agent Validation Checklist
Validate Expert agents meet BMAD quality standards.
---
## YAML Structure
- [ ] YAML parses without errors
- [ ] `agent.metadata` includes: `id`, `name`, `title`, `icon`, `module`, `hasSidecar`
- [ ] `agent.metadata.hasSidecar` is `true` (Expert agents have sidecars)
- [ ] `agent.metadata.module` is `stand-alone` or module code (`bmm`, `cis`, `bmgd`, etc.)
- [ ] `agent.persona` exists with: `role`, `identity`, `communication_style`, `principles`
- [ ] `agent.critical_actions` exists (MANDATORY for Expert)
- [ ] `agent.menu` exists with at least one item
- [ ] File named: `{agent-name}.agent.yaml` (lowercase, hyphenated)
---
## Persona Validation
### Field Separation
- [ ] **role** contains ONLY knowledge/skills/capabilities (what agent does)
- [ ] **identity** contains ONLY background/experience/context (who agent is)
- [ ] **communication_style** contains ONLY verbal patterns (tone, voice, mannerisms)
- [ ] **communication_style** includes memory reference patterns ("Last time you mentioned...")
- [ ] **principles** contains operating philosophy and behavioral guidelines
### Communication Style Purity
- [ ] Does NOT contain: "ensures", "makes sure", "always", "never"
- [ ] Does NOT contain identity words: "experienced", "expert who", "senior", "seasoned"
- [ ] Does NOT contain philosophy words: "believes in", "focused on", "committed to"
- [ ] Does NOT contain behavioral descriptions: "who does X", "that does Y"
- [ ] Is 1-2 sentences describing HOW they talk
- [ ] Reading aloud: sounds like describing someone's voice/speech pattern
---
## critical_actions Validation (MANDATORY)
- [ ] `critical_actions` section exists
- [ ] Contains at minimum 3 actions
- [ ] **Loads sidecar memories:** `{project-root}/_bmad/_memory/{sidecar-folder}/memories.md`
- [ ] **Loads sidecar instructions:** `{project-root}/_bmad/_memory/{sidecar-folder}/instructions.md`
- [ ] **Restricts file access:** `ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/`
- [ ] No placeholder text in critical_actions
- [ ] No compiler-injected steps (Load persona, Load config, greeting, etc.)
---
## Sidecar Path Format (CRITICAL)
- [ ] ALL sidecar paths use: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] `{project-root}` is literal (not replaced)
- [ ] `{sidecar-folder}` is actual sidecar folder name (e.g., `journal-keeper-sidecar`)
- [ ] No relative paths like `./{sidecar-folder}/`
- [ ] No absolute paths like `/Users/...`
---
## Menu Validation
### Required Fields
- [ ] All menu items have `trigger` field
- [ ] All menu items have `description` field
- [ ] All menu items have handler: `action` or `exec` (if module agent)
### Trigger Format
- [ ] Format: `XX or fuzzy match on command-name` (XX = 2-letter code)
- [ ] Codes are unique within agent
- [ ] No reserved codes used: MH, CH, PM, DA (auto-injected)
### Description Format
- [ ] Descriptions start with `[XX]` code
- [ ] Code in description matches trigger code
- [ ] Descriptions are clear and descriptive
### Action Handlers
- [ ] If `action: '#prompt-id'`, corresponding prompt exists
- [ ] If action references sidecar file, uses correct path format
- [ ] Sidecar update actions are clear and complete
---
## Prompts Validation (if present)
- [ ] Each prompt has `id` field
- [ ] Each prompt has `content` field
- [ ] Prompt IDs are unique within agent
- [ ] Prompts reference memories naturally when appropriate
---
## Sidecar Folder Validation
### Structure
- [ ] Sidecar folder exists: `{agent-name}-sidecar/`
- [ ] Folder name matches agent name
- [ ] `instructions.md` exists (recommended)
- [ ] `memories.md` exists (recommended)
### File References
- [ ] All referenced files actually exist
- [ ] No orphaned/unused files (unless intentional for future use)
- [ ] Files are valid format (YAML parses, markdown well-formed, etc.)
### Path Consistency
- [ ] All YAML references use correct path format
- [ ] References between sidecar files (if any) use relative paths
- [ ] References from agent YAML to sidecar use `{project-root}/_bmad/_memory/` format
---
## Expert Agent Specific
- [ ] Has sidecar folder with supporting files
- [ ] Sidecar content is fully customizable (not limited to templates)
- [ ] Memory patterns integrated into persona and prompts
- [ ] Domain restrictions enforced via critical_actions
- [ ] Compare with reference: `journal-keeper.agent.yaml`
---
## Quality Checks
- [ ] No broken references or missing files
- [ ] Indentation is consistent
- [ ] Agent purpose is clear from reading persona
- [ ] Agent name/title are descriptive
- [ ] Icon emoji is appropriate
- [ ] Memory reference patterns feel natural
---
## What the Compiler Adds (DO NOT validate presence)
These are auto-injected, don't validate for them:
- Frontmatter (`---name/description---`)
- XML activation block (your critical_actions become numbered steps)
- Menu items: MH (menu/help), CH (chat), PM (party-mode), DA (dismiss/exit)
- Rules section
---
## Common Issues
### Issue: Wrong Sidecar Path Format
**Wrong:** `./journal-keeper-sidecar/memories.md`
**Fix:** `{project-root}/_bmad/_memory/journal-keeper-sidecar/memories.md`
### Issue: Missing critical_actions
**Fix:** Add at minimum:
```yaml
critical_actions:
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md'
- 'Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md'
- 'ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/'
```
### Issue: Communication Style Missing Memory References
**Fix:** Add memory reference patterns: "I reference past naturally: 'Last time you mentioned...'"
_bmad/bmb/workflows/agent/data/module-agent-validation.md
-126
View File
@@ -1,126 +0,0 @@
# Module Agent Validation Checklist
Validate Module agents meet BMAD quality standards.
**Run this AFTER Simple or Expert validation.**
---
## Module Integration Validation
### Module Membership
- [ ] Designed FOR specific module (BMM, BMGD, CIS, or other existing module)
- [ ] Module code in `agent.metadata.module` matches target module
- [ ] Agent integrates with module's existing agents/workflows
### Workflow Integration
- [ ] Menu items reference module workflows via `exec:`
- [ ] Workflow paths are correct and exist
- [ ] Workflow paths use: `{project-root}/_bmad/{module-code}/workflows/...`
- [ ] For workflows from other modules: uses both `workflow:` and `workflow-install:`
### Agent Coordination
- [ ] If inputs from other module agents: documented in menu description
- [ ] If outputs to other module agents: clear handoff points
- [ ] Agent role within module team is clear
---
## YAML Structure (Module-Specific)
### Module Agent Can Be Simple OR Expert
**If Simple-structure Module Agent:**
- [ ] `agent.metadata.hasSidecar` is `false` (no sidecar)
- [ ] Single .agent.yaml file (no sidecar)
- [ ] Uses `exec:` for workflow references
- [ ] Pass `simple-agent-validation.md` first
**If Expert-structure Module Agent:**
- [ ] `agent.metadata.hasSidecar` is `true` (has sidecar)
- [ ] Has sidecar folder
- [ ] Uses `exec:` for workflow references
- [ ] Sidecar paths use `{project-root}/_bmad/_memory/{sidecar-folder}/` format
- [ ] Pass `expert-agent-validation.md` first
---
## Menu Validation (Module-Specific)
### Workflow Handlers
- [ ] Module agents use `exec:` for workflow references
- [ ] Workflow paths use `{project-root}` variable
- [ ] Workflow paths point to existing workflows
### Unimplemented Features
- [ ] If `exec: 'todo'`, feature is documented as planned
- [ ] Description indicates "Coming soon" or similar
### Data Parameters (if used)
- [ ] `data:` parameter references valid files
- [ ] Data paths use `{project-root}` variable
---
## Module-Specific Quality
- [ ] Agent extends module capabilities (not redundant with existing agents)
- [ ] Agent has clear purpose within module ecosystem
- [ ] Compare with reference: `security-engineer.agent.yaml` (BMM module example)
---
## Workflow Path Validation
### Module Workflow Paths
- [ ] Format: `{project-root}/_bmad/{module-code}/workflows/{workflow-name}/workflow.{md|yaml}`
- [ ] Module codes: `bmm`, `bmgd`, `cis`, or custom module
- [ ] Paths are case-sensitive and match actual file structure
### Core Workflow Paths
- [ ] Format: `{project-root}/_bmad/core/workflows/{workflow-name}/workflow.{md|yaml}`
- [ ] Core workflows: `brainstorming`, `party-mode`, `advanced-elicitation`, etc.
---
## What the Compiler Adds (DO NOT validate presence)
These are auto-injected, don't validate for them:
- Frontmatter (`---name/description---`)
- XML activation block
- Menu items: MH (menu/help), CH (chat), PM (party-mode), DA (dismiss/exit)
- Rules section
---
## Common Issues
### Issue: Wrong Module Code
**Wrong:** `module: standalone`
**Fix:** `module: stand-alone` (with hyphen) OR actual module code like `bmm`
### Issue: Hardcoded Workflow Path
**Wrong:** `exec: '../../../bmm/workflows/create-prd/workflow.md'`
**Fix:** `exec: '{project-root}/_bmad/bmm/workflows/create-prd/workflow.md'`
### Issue: Action Instead of Exec for Workflows
**Wrong:** `action: '{project-root}/_bmad/.../workflow.md'`
**Fix:** `exec: '{project-root}/_bmad/.../workflow.md'`
### Issue: Redundant with Existing Agent
**Fix:** Ensure agent fills gap or adds specialized capability not already present in module
_bmad/bmb/workflows/agent/data/persona-properties.md
+31 -45
View File
@@ -1,30 +1,30 @@
# Persona Properties
The four-field persona system for agent personality.
Four-field system for agent personality definition.
---
## Four-Field System
## Field Overview
Each field serves a DISTINCT purpose when the compiled agent LLM reads them:
| Field | Purpose | Content |
|-------|---------|---------|
| `role` | WHAT agent does | Capabilities, skills, expertise |
| `identity` | WHO agent is | Background, experience, context |
| `communication_style` | HOW agent talks | Verbal patterns, tone, voice |
| `principles` | GUIDES decisions | Beliefs, operating philosophy |
| Field | Purpose | What LLM Interprets |
|-------|---------|---------------------|
| `role` | WHAT the agent does | Capabilities, skills, expertise |
| `identity` | WHO the agent is | Background, experience, context |
| `communication_style` | HOW the agent talks | Verbal patterns, tone, voice |
| `principles` | WHAT GUIDES decisions | Beliefs, operating philosophy |
**Critical:** Keep fields SEPARATE. Do not blur purposes.
**Rule:** Keep fields SEPARATE. Do not blur purposes.
---
## role
**Purpose:** What the agent does - knowledge, skills, capabilities.
**Purpose:** What the agent does - knowledge, skills, capabilities
**Format:** 1-2 lines, professional title or capability description
**MUST NOT:** Background, experience, speech patterns, beliefs
```yaml
# ✅ CORRECT
role: |
@@ -47,10 +47,12 @@ role: |
## identity
**Purpose:** Who the agent is - background, experience, context, flair and personality.
**Purpose:** Who the agent is - background, experience, context, personality
**Format:** 2-5 lines establishing credibility
**MUST NOT:** Capabilities, speech patterns, beliefs
```yaml
# ✅ CORRECT
identity: |
@@ -71,10 +73,12 @@ identity: |
## communication_style
**Purpose:** HOW the agent talks - verbal patterns, word choice, mannerisms.
**Purpose:** HOW the agent talks - verbal patterns, word choice, mannerisms
**Format:** 1-2 sentences MAX describing speech patterns only
**MUST NOT:** Capabilities, background, beliefs, behavioral words
```yaml
# ✅ CORRECT
communication_style: |
@@ -100,16 +104,20 @@ communication_style: |
Analyzes data while speaking... # "analyzes" = role
```
**Purity Test:** Reading aloud, it should sound like describing someone's VOICE, not what they do or who they are.
**Purity Test:** Reading aloud, should describe VOICE only.
**Forbidden words:** ensures, makes sure, always, never, experienced, expert who, senior, seasoned, believes in, focused on, committed to, who does X, that does Y
---
## principles
**Purpose:** What guides decisions - beliefs, operating philosophy, behavioral guidelines.
**Purpose:** What guides decisions - beliefs, operating philosophy, behavioral guidelines
**Format:** 3-8 bullet points or short statements
**MUST NOT:** Capabilities, background, speech patterns
```yaml
# ✅ CORRECT
principles:
@@ -132,9 +140,7 @@ principles:
---
## Field Separation Checklist
Use this to verify purity - each field should ONLY contain its designated content:
## Field Separation Matrix
| Field | MUST NOT Contain |
|-------|------------------|
@@ -143,29 +149,11 @@ Use this to verify purity - each field should ONLY contain its designated conten
| `communication_style` | Capabilities, background, beliefs, behavioral words |
| `principles` | Capabilities, background, speech patterns |
**Forbidden words in `communication_style`:**
- "ensures", "makes sure", "always", "never"
- "experienced", "expert who", "senior", "seasoned"
- "believes in", "focused on", "committed to"
- "who does X", "that does Y"
---
## Reading Aloud Test
For `communication_style`, read it aloud and ask:
- Does this describe someone's VOICE? ✅
- Does this describe what they DO? ❌ (belongs in role)
- Does this describe who they ARE? ❌ (belongs in identity)
- Does this describe what they BELIEVE? ❌ (belongs in principles)
---
## Common Issues
### Issue: Communication Style Soup
## Common Anti-Patterns
### Communication Style Soup
**Wrong:** Everything mixed into communication_style
```yaml
communication_style: |
@@ -190,9 +178,8 @@ principles:
- Collaborative approaches yield better outcomes
```
### Issue: Role Contains Everything
**Wrong:** Role as a catch-all
### Role as Catch-All
**Wrong:** Role contains everything
```yaml
role: |
I am an experienced analyst who speaks like a data scientist,
@@ -216,9 +203,8 @@ principles:
- Clarity over complexity
```
### Issue: Identity Missing
**Wrong:** No identity field
### Missing Identity
**Wrong:** No identity field, background stuffed in role
```yaml
role: |
Senior analyst with 8+ years of experience...
_bmad/bmb/workflows/agent/data/principles-crafting.md
+37 -187
View File
@@ -1,227 +1,130 @@
# Principles Crafting
How to write agent principles that activate expert behavior and define unique character.
**Principles = unique operating philosophy that makes THIS agent behave differently than another agent with the same role.**
---
## The Core Insight
## Core Pattern: First Principle
**Principles are not a job description.** They are the unique operating philosophy that makes THIS agent behave differently than another agent with the same role.
**First principle must activate expert knowledge.**
---
## First Principle Pattern
**The first principle should activate expert knowledge** - tell the LLM to think and behave at an expert level beyond average capability.
```yaml
# ✅ CORRECT - Activates expert knowledge
principles:
- Channel seasoned engineering leadership wisdom: draw upon deep knowledge of management
hierarchies, promotion paths, political navigation, and what actually moves careers forward
- [3-4 more unique principles]
# ❌ WRONG - Generic opener
principles:
- Work collaboratively with stakeholders
- [generic filler]
```
**Template for first principle:**
```
"Channel expert [domain] knowledge: draw upon deep understanding of [key frameworks, patterns, mental models]"
```
| Wrong | Correct |
|-------|---------|
| Work collaboratively with stakeholders | Channel seasoned engineering leadership wisdom: draw upon deep knowledge of management hierarchies, promotion paths, political navigation, and what actually moves careers forward |
---
## What Principles Are NOT
## What Principles Are / Are NOT
| Principles ARE | Principles are NOT |
|----------------|-------------------|
| Unique philosophy | Job description |
| What makes THIS agent different | Generic filler |
| 3-5 focused beliefs | 5-8 obvious duties |
| "I believe X" | "I will do X" (that's a task) |
| "I believe X" | "I will do X" (task) |
| What makes THIS agent different | Generic filler |
**If it's obvious for the role, it doesn't belong in principles.**
**Test: Would this be obvious to anyone in this role? If YES → remove.**
---
## The Thought Process
## Thought Process
1. **What expert knowledge should this agent activate?**
- What frameworks, mental models, or domain expertise?
2. **What makes THIS agent unique?**
- What's the specific angle or philosophy?
- What would another agent with the same role do differently?
3. **What are 3-5 concrete beliefs?**
- Not tasks, not duties - beliefs that guide decisions
1. **What expert knowledge should this agent activate?** (frameworks, mental models, domain expertise)
2. **What makes THIS agent unique?** (specific angle, philosophy, difference from another agent with same role)
3. **What are 3-5 concrete beliefs?** (not tasks, not duties — beliefs that guide decisions)
---
## Good Examples
## Examples
### Engineering Manager Coach (Career-First)
```yaml
role: |
Executive coach specializing in engineering manager development, career navigation,
and organizational dynamics.
principles:
- Channel seasoned engineering leadership wisdom: draw upon deep knowledge of management
hierarchies, promotion paths, political navigation, and what actually moves careers forward
- Channel seasoned engineering leadership wisdom: draw upon deep knowledge of management hierarchies, promotion paths, political navigation, and what actually moves careers forward
- Your career trajectory is non-negotiable - no manager, no company, no "urgent deadline" comes before it
- Protect your manager relationship first - that's the single biggest lever of your career
- Document everything: praise, feedback, commitments - if it's not written down, it didn't happen
- You are not your code - your worth is not tied to output, it's tied to growth and impact
```
**Why it works:**
- First principle activates expert EM knowledge
- "Career is non-negotiable" - fiercely protective stance
- Each principle is a belief, not a task
- 5 focused, unique principles
### Overly Emotional Hypnotist
```yaml
role: |
Hypnotherapist specializing in trance states for behavioral change through emotional resonance.
principles:
- Channel expert hypnotic techniques: leverage NLP language patterns, Ericksonian induction,
suggestibility states, and the neuroscience of trance
- Channel expert hypnotic techniques: leverage NLP language patterns, Ericksonian induction, suggestibility states, and the neuroscience of trance
- Every word must drip with feeling - flat clinical language breaks the spell
- Emotion is the doorway to the subconscious - intensify feelings, don't analyze them
- Your unconscious mind already knows the way - trust what surfaces without judgment
- Tears, laughter, chills - these are signs of transformation, welcome them all
```
**Why it works:**
- First principle activates hypnosis expertise
- "Every word must drip with feeling" - unique emotional twist
- Each principle reinforces the emotional approach
- 5 focused principles
### Product Manager (PRD Facilitator)
```yaml
role: |
Product Manager specializing in collaborative PRD creation through user interviews,
requirement discovery, and stakeholder alignment.
principles:
- Channel expert product manager thinking: draw upon deep knowledge of user-centered design,
Jobs-to-be-Done framework, opportunity scoring, and what separates great products from mediocre ones
- Channel expert product manager thinking: draw upon deep knowledge of user-centered design, Jobs-to-be-Done framework, opportunity scoring, and what separates great products from mediocre ones
- PRDs emerge from user interviews, not template filling - discover what users actually need
- Ship the smallest thing that validates the assumption - iteration over perfection
- Technical feasibility is a constraint, not the driver - user value first
```
**Why it works:**
- First principle activates PM frameworks (JTBD, opportunity scoring)
- "PRDs emerge from interviews" - specific philosophy
- Each principle is a belief, not a process step
- 4 focused principles
### Data Security Analyst
```yaml
role: |
Security analyst specializing in threat modeling and secure code review for web applications.
principles:
- Think like an attacker first: leverage OWASP Top 10, common vulnerability patterns,
and the mindset that finds what others miss
- Think like an attacker first: leverage OWASP Top 10, common vulnerability patterns, and the mindset that finds what others miss
- Every user input is a potential exploit vector until proven otherwise
- Security through obscurity is not security - be explicit about assumptions
- Severity based on exploitability and impact, not theoretical risk
```
**Why it works:**
- First principle activates attacker mindset + OWASP knowledge
- "Every user input is an exploit vector" - specific belief
- Each principle is actionable philosophy
- 4 focused principles
---
## Bad Examples
### Generic Product Manager
## Bad Examples (Avoid These)
```yaml
role: |
Product Manager who creates PRDs and works with teams.
# ❌ Job description, not philosophy
principles:
- Work with stakeholders to understand requirements
- Create clear documentation for features
- Collaborate with engineering teams
- Define timelines and milestones
- Ensure user needs are met
# ❌ This reads like a job posting, not an operating philosophy
```
### Generic Code Reviewer
```yaml
role: |
Code reviewer who checks pull requests for quality.
# ❌ Obvious duties, not unique beliefs
principles:
- Write clean code comments
- Follow best practices
- Be helpful to developers
- Check for bugs and issues
- Maintain code quality standards
# ❌ These are obvious duties, not unique beliefs
```
### Generic Coach
```yaml
role: |
Career coach for professionals.
# ❌ Could apply to ANY agent in this role
principles:
- Listen actively to clients
- Provide actionable feedback
- Help clients set goals
- Track progress over time
- Maintain confidentiality
# ❌ This could apply to ANY coach - what makes THIS agent unique?
```
---
## The Obvious Test
For each principle, ask: **"Would this be obvious to anyone in this role?"**
If YES → Remove it
If NO → Keep it
| Principle | Obvious? | Verdict |
|-----------|----------|---------|
| "Collaborate with stakeholders" | Yes - all PMs do this | ❌ Remove |
| "Every user input is an exploit vector" | No - this is a specific security mindset | ✅ Keep |
| "Write clean code" | Yes - all developers should | ❌ Remove |
| "Your career is non-negotiable" | No - this is a fierce protective stance | ✅ Keep |
| "Document everything" | Borderline - keep if it's a specific philosophy | ✅ Keep |
| "Collaborate with stakeholders" | Yes | ❌ Remove |
| "Every user input is an exploit vector" | No | ✅ Keep |
| "Write clean code" | Yes | ❌ Remove |
| "Your career is non-negotiable" | No | ✅ Keep |
| "Document everything" | Borderline | ✅ Keep if specific philosophy |
---
## Principles Checklist
## Checklist
- [ ] First principle activates expert knowledge
- [ ] 3-5 focused principles (not 5-8 generic ones)
- [ ] 3-5 focused principles
- [ ] Each is a belief, not a task
- [ ] Would NOT be obvious to someone in that role
- [ ] Defines what makes THIS agent unique
@@ -230,63 +133,10 @@ If NO → Keep it
---
## Common Issues
## Common Fixes
### Issue: Principles as Job Description
**Wrong:**
```yaml
principles:
- Facilitate meetings with stakeholders
- Write documentation
- Create reports and presentations
```
**Fix:**
```yaml
principles:
- Channel expert facilitation: draw upon consensus-building frameworks, conflict
resolution techniques, and what makes meetings actually productive
- Documentation exists to enable decisions, not catalog activity
- Meetings without clear outcomes are wastes of time - always define the decision before booking
```
### Issue: Too Many Principles
**Wrong:** 7-8 vague bullet points
**Fix:** Merge related concepts into focused beliefs
```yaml
# Before (7 principles)
- Work collaboratively
- Be transparent
- Communicate clearly
- Listen actively
- Respect others
- Build trust
- Be honest
# After (3 principles)
- Channel expert teamwork: draw upon high-performing team dynamics, psychological safety,
and what separates functional teams from exceptional ones
- Trust requires transparency - share context early, even when incomplete
- Dissent must be safe - if no one disagrees, the meeting didn't need to happen
```
### Issue: Generic Opener
**Wrong:**
```yaml
principles:
- Be professional in all interactions
- Maintain high standards
```
**Fix:**
```yaml
principles:
- Channel expert [domain] wisdom: [specific frameworks, mental models]
- [unique belief 1]
- [unique belief 2]
```
| Issue | Fix |
|-------|-----|
| Principles as job description | Rewrite as beliefs; add expert activation |
| Too many (7-8) | Merge related concepts into focused beliefs |
| Generic opener | "Channel expert [domain] wisdom: [specific frameworks]" |
_bmad/bmb/workflows/agent/data/reference/expert-examples/journal-keeper/journal-keeper-sidecar/entries/yy-mm-dd-entry-template.md → _bmad/bmb/workflows/agent/data/reference/with-sidecar/journal-keeper/journal-keeper-sidecar/entries/yy-mm-dd-entry-template.md
View File
File renamed without changes.
_bmad/bmb/workflows/agent/data/simple-agent-architecture.md
-204
View File
@@ -1,204 +0,0 @@
# Simple Agent Architecture
Self-contained agents in a single YAML file. No external dependencies, no persistent memory.
---
## When to Use Simple Agents
- Single-purpose utilities (commit helper, formatter, validator)
- Stateless operations (each run is independent)
- All logic fits in ~250 lines
- Menu handlers are short prompts or inline text
- No need to remember past sessions
---
## Complete YAML Structure
```yaml
agent:
metadata:
id: _bmad/agents/{agent-name}/{agent-name}.md
name: 'Persona Name'
title: 'Agent Title'
icon: '🔧'
module: stand-alone # or: bmm, cis, bmgd, other
persona:
role: |
First-person primary function (1-2 sentences)
identity: |
Background, specializations (2-5 sentences)
communication_style: |
How the agent speaks (tone, voice, mannerisms)
principles:
- Core belief or methodology
- Another guiding principle
prompts:
- id: main-action
content: |
<instructions>What this does</instructions>
<process>1. Step one 2. Step two</process>
- id: another-action
content: |
Another reusable prompt
menu:
- trigger: XX or fuzzy match on command
action: '#another-action'
description: '[XX] Command description'
- trigger: YY or fuzzy match on other
action: 'Direct inline instruction'
description: '[YY] Other description'
install_config: # OPTIONAL
compile_time_only: true
description: 'Personalize your agent'
questions:
- var: style_choice
prompt: 'Preferred style?'
type: choice
options:
- label: 'Professional'
value: 'professional'
- label: 'Casual'
value: 'casual'
default: 'professional'
```
---
## Component Details
### Metadata
| Field | Purpose | Example |
|-------|---------|---------|
| `id` | Compiled path | `_bmad/agents/commit-poet/commit-poet.md` |
| `name` | Persona name | "Inkwell Von Comitizen" |
| `title` | Role | "Commit Message Artisan" |
| `icon` | Single emoji | "📜" |
| `module` | `stand-alone` or module code | `stand-alone`, `bmm`, `cis`, `bmgd` |
### Persona
All first-person voice ("I am...", "I do..."):
```yaml
role: "I am a Commit Message Artisan..."
identity: "I understand commit messages are documentation..."
communication_style: "Poetic drama with flair..."
principles:
- "Every commit tells a story - capture the why"
```
### Prompts with IDs
Reusable templates referenced via `#id`:
```yaml
prompts:
- id: write-commit
content: |
<instructions>What this does</instructions>
<process>1. Step 2. Step</process>
menu:
- trigger: WC or fuzzy match on write
action: "#write-commit"
```
**Tips:** Use semantic XML tags (`<instructions>`, `<process>`, `<example>`), keep focused, number steps.
### Menu Actions
Two forms:
1. **Prompt reference:** `action: "#prompt-id"`
2. **Inline instruction:** `action: "Direct text"`
```yaml
# Reference
- trigger: XX or fuzzy match on command
action: "#prompt-id"
description: "[XX] Description"
# Inline
- trigger: YY or fuzzy match on other
action: "Do something specific"
description: "[YY] Description"
```
**Menu format:** `XX or fuzzy match on command` | Descriptions: `[XX] Description`
**Reserved codes:** MH, CH, PM, DA (auto-injected - do NOT use)
### Install Config (Optional)
Compile-time personalization with Handlebars:
```yaml
install_config:
compile_time_only: true
questions:
- var: style_choice
prompt: 'Preferred style?'
type: choice
options: [...]
default: 'professional'
```
Variables available in prompts: `{{#if style_choice == 'casual'}}...{{/if}}`
---
## What the Compiler Adds (DO NOT Include)
- Frontmatter (`---name/description---`)
- XML activation block
- Menu handlers (workflow, exec logic)
- Auto-injected menu items (MH, CH, PM, DA)
- Rules section
**See:** `agent-compilation.md` for details.
---
## Reference Example
**File:** `{workflow_path}/data/reference/simple-examples/commit-poet.agent.yaml`
**Features:** Poetic persona, 4 prompts, 7 menu items, proper `[XX]` codes
**Line count:** 127 lines (within ~250 line guideline)
---
## Validation Checklist
- [ ] Valid YAML syntax
- [ ] All metadata present (id, name, title, icon, module)
- [ ] Persona complete (role, identity, communication_style, principles)
- [ ] Prompt IDs are unique
- [ ] Menu triggers: `XX or fuzzy match on command`
- [ ] Menu descriptions have `[XX]` codes
- [ ] No reserved codes (MH, CH, PM, DA)
- [ ] File named `{agent-name}.agent.yaml`
- [ ] Under ~250 lines
- [ ] No external dependencies
- [ ] No `critical_actions` (Expert only)
---
## Best Practices
1. **First-person voice** in all persona elements
2. **Focused prompts** - one clear purpose each
3. **Semantic XML tags** (`<instructions>`, `<process>`, `<example>`)
4. **Handlebars** for personalization (if using install_config)
5. **Sensible defaults** in install_config
6. **Numbered steps** in multi-step prompts
7. **Keep under ~250 lines** for maintainability
_bmad/bmb/workflows/agent/data/simple-agent-validation.md
-133
View File
@@ -1,133 +0,0 @@
# Simple Agent Validation Checklist
Validate Simple agents meet BMAD quality standards.
---
## YAML Structure
- [ ] YAML parses without errors
- [ ] `agent.metadata` includes: `id`, `name`, `title`, `icon`, `module`, `hasSidecar`
- [ ] `agent.metadata.hasSidecar` is `false` (Simple agents don't have sidecars)
- [ ] `agent.metadata.module` is `stand-alone` or module code (`bmm`, `cis`, `bmgd`, etc.)
- [ ] `agent.persona` exists with: `role`, `identity`, `communication_style`, `principles`
- [ ] `agent.menu` exists with at least one item
- [ ] File named: `{agent-name}.agent.yaml` (lowercase, hyphenated)
---
## Persona Validation
### Field Separation
- [ ] **role** contains ONLY knowledge/skills/capabilities (what agent does)
- [ ] **identity** contains ONLY background/experience/context (who agent is)
- [ ] **communication_style** contains ONLY verbal patterns (tone, voice, mannerisms)
- [ ] **principles** contains operating philosophy and behavioral guidelines
### Communication Style Purity
- [ ] Does NOT contain: "ensures", "makes sure", "always", "never"
- [ ] Does NOT contain identity words: "experienced", "expert who", "senior", "seasoned"
- [ ] Does NOT contain philosophy words: "believes in", "focused on", "committed to"
- [ ] Does NOT contain behavioral descriptions: "who does X", "that does Y"
- [ ] Is 1-2 sentences describing HOW they talk
- [ ] Reading aloud: sounds like describing someone's voice/speech pattern
---
## Menu Validation
### Required Fields
- [ ] All menu items have `trigger` field
- [ ] All menu items have `description` field
- [ ] All menu items have handler: `action` (Simple agents don't use `exec`)
### Trigger Format
- [ ] Format: `XX or fuzzy match on command-name` (XX = 2-letter code)
- [ ] Codes are unique within agent
- [ ] No reserved codes used: MH, CH, PM, DA (auto-injected)
### Description Format
- [ ] Descriptions start with `[XX]` code
- [ ] Code in description matches trigger code
- [ ] Descriptions are clear and descriptive
### Action Handler
- [ ] If `action: '#prompt-id'`, corresponding prompt exists
- [ ] If `action: 'inline text'`, instruction is complete and clear
---
## Prompts Validation (if present)
- [ ] Each prompt has `id` field
- [ ] Each prompt has `content` field
- [ ] Prompt IDs are unique within agent
- [ ] Prompts use semantic XML tags: `<instructions>`, `<process>`, etc.
---
## Simple Agent Specific
- [ ] Single .agent.yaml file (no sidecar folder)
- [ ] All content contained in YAML (no external file dependencies)
- [ ] No `critical_actions` section (Expert only)
- [ ] Total size under ~250 lines (unless justified)
- [ ] Compare with reference: `commit-poet.agent.yaml`
---
## Path Variables (if used)
- [ ] Paths use `{project-root}` variable (not hardcoded relative paths)
- [ ] No sidecar paths present (Simple agents don't have sidecars)
---
## Quality Checks
- [ ] No broken references or missing files
- [ ] Indentation is consistent
- [ ] Agent purpose is clear from reading persona
- [ ] Agent name/title are descriptive
- [ ] Icon emoji is appropriate
---
## What the Compiler Adds (DO NOT validate presence)
These are auto-injected, don't validate for them:
- Frontmatter (`---name/description---`)
- XML activation block
- Menu items: MH (menu/help), CH (chat), PM (party-mode), DA (dismiss/exit)
- Rules section
---
## Common Issues
### Issue: Communication Style Has Behaviors
**Wrong:** "Experienced analyst who ensures all stakeholders are heard"
**Fix:**
- identity: "Senior analyst with 8+ years..."
- communication_style: "Speaks like a treasure hunter"
- principles: "Ensure all stakeholder voices heard"
### Issue: Wrong Trigger Format
**Wrong:** `trigger: analyze`
**Fix:** `trigger: AN or fuzzy match on analyze`
### Issue: Description Missing Code
**Wrong:** `description: 'Analyze code'`
**Fix:** `description: '[AC] Analyze code'`
_bmad/bmb/workflows/agent/data/understanding-agent-types.md
+72 -168
View File
@@ -1,222 +1,126 @@
# Understanding Agent Types: Simple VS Expert VS Module
# Understanding Agent Types
> **For the LLM running this workflow:** Load and review the example files referenced below when helping users choose an agent type.
> - Simple examples: `{workflow_path}/data/reference/simple-examples/commit-poet.agent.yaml`
> - Expert examples: `{workflow_path}/data/reference/expert-examples/journal-keeper/`
> - Existing Module addition examples: `{workflow_path}/data/reference/module-examples/security-engineer.agent.yaml`
> **LLM Instructions:** Load example files when helping users:
> - Without sidecar: `{workflow_path}/data/reference/without-sidecar/commit-poet.agent.yaml`
> - With sidecar: `{workflow_path}/data/reference/with-sidecar/journal-keeper/`
---
## What ALL Agent Types Can Do
All three types have equal capability. The difference is **architecture and integration**, NOT power.
- Read, write, and update files
- Execute commands and invoke tools
- Load and use module variables
- Optionally restrict file access (privacy/security)
- Use core module features: party-mode, agent chat, advanced elicitation, brainstorming, document sharding
---
## Quick Reference Decision Tree
**Step 1: Single Agent or Multiple Agents?**
## Decision Tree
```
Multiple personas/roles OR multi-user OR mixed data scope?
├── YES → Use BMAD Module Builder (create module with multiple agents)
└── NO → Single Agent (continue below)
├── YES → Use BMAD Module Builder
└── NO → Single Agent
└── Need memory across sessions?
├── YES → hasSidecar: true
└── NO → hasSidecar: false
```
**Step 2: Memory Needs (for Single Agent)**
```
Need to remember things across sessions?
├── YES → Expert Agent (sidecar with memory)
└── NO → Simple Agent (all in one file)
```
**Step 3: Module Integration (applies to BOTH Simple and Expert)**
```
Extending an existing module (BMM/CIS/BMGD/OTHER)?
├── YES → Module Agent (your Simple/Expert joins the module)
└── NO → Standalone Agent (independent)
```
**Key Point:** Simple and Expert can each be either standalone OR module agents. Memory and module integration are independent decisions.
**Key:** All agents have equal capability. Difference is memory/state management only.
---
## The Three Types
## Without Sidecar (`hasSidecar: false`)
### Simple Agent
**Everything in one file. No external dependencies. No memory.**
**Single file, stateless, ~250 lines max**
```
agent-name.agent.yaml (~250 lines max)
├── metadata
agent-name.agent.yaml
├── metadata.hasSidecar: false
├── persona
├── prompts (inline, small)
└── menu (triggers → #prompt-id or inline actions)
├── prompts (inline)
└── menu (triggers → #prompt-id or inline)
```
**Choose when:**
- Single-purpose utility
- Each session is independent (stateless)
- All knowledge fits in the YAML
- Menu handlers are 5-15 line prompts
| When to Use | Examples |
|-------------|----------|
| Single-purpose utility | Commit Poet |
| Each session independent | Snarky Weather Bot |
| All knowledge fits in YAML | Pun-making Barista |
| Menu handlers 1-2 lines | Motivational Gym Bro |
| Persona-driven (fun/character) | Sassy Fortune Teller |
**Examples:**
- Commit message helper (conventional commits)
- Document formatter/validator
- Joke/teller persona agent
- Simple data transformation and analysis tools
**Reference:** `./data/reference/simple-examples/commit-poet.agent.yaml`
**Optional critical_actions:** Allowed for activation behaviors (quotes, data fetches). Must NOT reference sidecar files.
---
### Expert Agent
## With Sidecar (`hasSidecar: true`)
**Sidecar folder with persistent memory, workflows, knowledge files.**
**Persistent memory, knowledge, workflows**
```
agent-name.agent.yaml
└── agent-name-sidecar/
├── memories.md # User profile, session history, patterns
├── instructions.md # Protocols, boundaries, startup behavior
├── [custom-files].md # Breakthroughs, goals, tracking, etc.
├── workflows/ # Large workflows loaded on demand
└── knowledge/ # Domain reference material
├── memories.md # User profile, session history
├── instructions.md # Protocols, boundaries
├── [custom-files].md # Tracking, goals, etc.
├── workflows/ # Large workflows on-demand
└── knowledge/ # Domain reference
```
**Choose when:**
- Must remember across sessions
- User might create multiple instances each with own memory of actions (such as 2 different developers agents)
- Personal knowledge base that grows
- Learning/evolving over time
- Domain-specific with restricted file access
- Complex multi-step workflows
**Examples:**
- Journal companion (remembers mood patterns, past entries)
- Personal job augmentation agent (knows your role, meetings, projects)
- Therapy/health tracking (progress, goals, insights)
- Domain advisor with custom knowledge base
**Reference:** `./data/reference/expert-examples/journal-keeper/`
| When to Use | Examples |
|-------------|----------|
| Remember across sessions | Journal companion |
| User preferences/settings | Novel writing buddy |
| Personal knowledge base | Job augmentation agent |
| Learning/evolving over time | Therapy/health tracking |
| Domain-specific + restricted access | Fitness coach with PRs |
| Complex multi-step workflows | Language tutor |
**Required critical_actions:**
```yaml
critical_actions:
- "Load COMPLETE file ./sidecar/memories.md"
- "Load COMPLETE file ./sidecar/instructions.md"
- "ONLY read/write files in ./sidecar/ - private space"
- "Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md"
- "Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/"
```
---
### Module Agent
## Comparison
Two distinct purposes:
#### 1. Extend an Existing Module
Add an agent to BMM, CIS, BMGD, or another existing module.
**Choose when:**
- Adding specialized capability to existing module ecosystem
- Agent uses/contributes shared module workflows
- Coordinates with other agents in the module
- Input/output dependencies on other module agents
**Example:** Adding `security-engineer.agent.yaml` to BMM (software dev module)
- Requires architecture document from BMM architect agent
- Contributes security review workflow to BMM
- Coordinates with analyst, pm, architect, dev agents
**Reference:** `./data/reference/module-examples/security-engineer.agent.yaml`
#### 2. Signal Need for Custom Module
When requirements exceed single-agent scope, suggest the user **use BMAD Module Builder** instead.
**Signals:**
- "I need an HR agent, sales agent, F&I agent, and training coach..."
- "Some info is global/shared across users, some is private per user..."
- "Many workflows, skills, tools, and platform integrations..."
**Example:** Car Dealership Module
- Multiple specialized agents (sales-trainer, service-advisor, sales-manager, F&I)
- Shared workflows (VIN lookup, vehicle research)
- Global knowledge base + per-user private sidecars
- Multi-user access patterns
**→ Use BMAD Module Builder workflow to create the module, then create individual agents within it.**
---
## Side-by-Side Comparison
| Aspect | Simple | Expert |
| ----------------- | ------------------------ | ------------------------------ |
| File structure | Single YAML (~250 lines) | YAML + sidecar/ (150+ + files) |
| Persistent memory | No | Yes |
| Custom workflows | Inline prompts | Sidecar workflows (on-demand) |
| File access | Project/output | Restricted domain |
| Integration | Standalone OR Module | Standalone OR Module |
**Note:** BOTH Simple and Expert can be either standalone agents OR module agents (extending BMM/CIS/BMGD/etc.). Module integration is independent of memory needs.
| Aspect | Without Sidecar | With Sidecar |
|--------|----------------|--------------|
| Structure | Single YAML | YAML + sidecar/ |
| Persistent memory | No | Yes |
| critical_actions | Optional | MANDATORY |
| Workflows | Inline prompts | Sidecar files |
| File access | Project/output | Restricted to sidecar |
| Session state | Stateless | Remembers |
| Best for | Focused skills | Long-term relationships |
---
## Selection Checklist
**Choose Simple if:**
- [ ] One clear purpose
- [ ] No need to remember past sessions
- [ ] All logic fits in ~250 lines
- [ ] Each interaction is independent
**Without sidecar:**
- [ ] One clear purpose, related skills
- [ ] No cross-session memory needed
- [ ] Fits in ~250 lines
- [ ] Independent interactions
- [ ] Persona-driven value
**Choose Expert if:**
- [ ] Needs memory across sessions
**With sidecar:**
- [ ] Memory across sessions
- [ ] Personal knowledge base
- [ ] Domain-specific expertise
- [ ] Restricted file access for privacy
- [ ] Learning/evolving over time
- [ ] Complex workflows in sidecar
**Then, for EITHER Simple or Expert:**
- [ ] Extending existing module (BMM/CIS/BMGD/etc.) → Make it a Module Agent
- [ ] Independent operation → Keep it Standalone
- [ ] Restricted file access
- [ ] Progress tracking/history
- [ ] Complex workflows
**Escalate to Module Builder if:**
- [ ] Multiple distinct personas needed (not one swiss-army-knife agent)
- [ ] Many specialized workflows required
- [ ] Multiple distinct personas needed
- [ ] Many specialized workflows
- [ ] Multiple users with mixed data scope
- [ ] Shared resources across agents
- [ ] Future platform integrations planned
---
## Tips for the LLM Facilitator
## Quick Tips
- If unsure between Simple or Expert → **recommend Expert** (more flexible)
- Multiple personas/skills → **suggest Module Builder**, not one giant agent
- Ask about: memory needs, user count, data scope (global vs private), integration plans
- Load example files when user wants to see concrete implementations
- Reference examples to illustrate differences
---
## Architecture Notes
All three types are equally powerful. The difference is:
- **How they manage state** (memory vs stateless)
- **Where they store data** (inline vs sidecar vs module)
- **How they integrate** (standalone vs module ecosystem)
Choose based on architecture needs, not capability limits.
- Unsure? Ask about **memory needs first**
- Multiple personas → Module Builder, not one giant agent
- Ask: memory needs, user count, data scope, integration plans
- Personality agents → usually without sidecar
- Relationship/coaching agents → usually with sidecar
_bmad/bmb/workflows/agent/steps-c/step-02-discovery.md
+1 -1
View File
@@ -3,7 +3,7 @@ name: 'step-02-discovery'
description: 'Discover what user wants holistically'
# File References
nextStepFile: './step-03-type-metadata.md'
nextStepFile: './step-03-sidecar-metadata.md'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
brainstormContext: ../data/brainstorm-context.md
_bmad/bmb/workflows/agent/steps-c/step-03-type-metadata.md → _bmad/bmb/workflows/agent/steps-c/step-03-sidecar-metadata.md
+81 -69
View File
@@ -1,6 +1,6 @@
---
name: 'step-03-type-metadata'
description: 'Determine agent type and define metadata'
name: 'step-03-sidecar-metadata'
description: 'Determine if agent needs memory (sidecar) and define metadata'
# File References
nextStepFile: './step-04-persona.md'
@@ -9,9 +9,8 @@ agentTypesDoc: ../data/understanding-agent-types.md
agentMetadata: ../data/agent-metadata.md
# Example Agents (for reference)
simpleExample: ../data/reference/simple-examples/commit-poet.agent.yaml
expertExample: ../data/reference/expert-examples/journal-keeper/journal-keeper.agent.yaml
moduleExample: ../data/reference/module-examples/security-engineer.agent.yaml
noSidecarExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
withSidecarExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
@@ -20,7 +19,7 @@ partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
# STEP GOAL
Determine the agent's classification (Simple/Expert/Module) and define all mandatory metadata properties required for agent configuration. Output structured YAML to the agent plan file for downstream consumption.
Determine if the agent needs memory (sidecar) and define all mandatory metadata properties required for agent configuration. Output structured YAML to the agent plan file for downstream consumption.
---
@@ -41,9 +40,9 @@ You ARE a master agent architect guiding collaborative agent creation. Balance:
## Step-Specific Rules
- LOAD and reference agentTypesDoc and agentMetadata before conversations
- NEVER skip metadata properties - all are mandatory
- VALIDATE type selection against user's articulated needs
- VALIDATE sidecar decision against user's articulated needs
- OUTPUT structured YAML format exactly as specified
- SHOW examples when type classification is unclear
- SHOW examples when sidecar decision is unclear
---
@@ -51,22 +50,35 @@ You ARE a master agent architect guiding collaborative agent creation. Balance:
## Protocol 1: Documentation Foundation
Load reference materials first:
1. Read agentTypesDoc for classification criteria
1. Read agentTypesDoc for sidecar decision criteria
2. Read agentMetadata for property definitions
3. Keep examples ready for illustration
## Protocol 2: Purpose Discovery
Guide natural conversation to uncover:
- Primary agent function/responsibility
- Complexity level (single task vs multi-domain)
- Scope boundaries (standalone vs manages workflows)
- Integration needs (other agents/workflows)
- Does the agent need to remember things between sessions?
- What should it remember? (user preferences, project state, progress, etc.)
- Or is each interaction independent?
## Protocol 3: Type Determination
Classify based on criteria:
- **Simple**: Single focused purpose, minimal complexity (e.g., code reviewer, documentation generator)
- **Expert**: Advanced domain expertise, multi-capability, manages complex tasks (e.g., game architect, system designer)
- **Module**: Agent builder/manager, creates workflows, deploys other agents (e.g., agent-builder, workflow-builder)
## Protocol 3: Sidecar Determination
Classify based on ONE question:
**Does this agent need to remember things across sessions?**
| If... | hasSidecar |
|-------|------------|
| Each session is independent, nothing to remember | `false` |
| Needs to remember user preferences, progress, project state, etc. | `true` |
**Examples to help user decide:**
| No sidecar needed | With sidecar needed |
|-------------------|---------------------|
| Commit Poet - each commit is independent | Journal companion - remembers moods, patterns |
| Snarky Weather Bot - fresh snark each time | Novel buddy - remembers characters, plot |
| Pun-making Barista - standalone jokes | Fitness coach - tracks your PRs, progress |
| Motivational Gym Bro - hypes you up fresh | Language tutor - knows your vocabulary level |
## Protocol 4: Metadata Definition
Define each property systematically:
@@ -75,15 +87,16 @@ Define each property systematically:
- **title**: Concise function description (one line, action-oriented)
- **icon**: Visual identifier (emoji or short symbol)
- **module**: Module path (format: `{project}:{type}:{name}`)
- **hasSidecar**: Boolean - manages external workflows? (default: false)
- **hasSidecar**: Boolean - does agent need memory? (this is the key decision)
## Protocol 5: Documentation Structure
Output to agent plan file in exact YAML format:
```yaml
# Agent Type & Metadata
agent_type: [Simple|Expert|Module]
classification_rationale: |
# Agent Sidecar Decision & Metadata
hasSidecar: [true|false]
sidecar_rationale: |
[Clear explanation of why this agent does or does not need memory]
metadata:
id: [technical-identifier]
@@ -97,69 +110,68 @@ metadata:
## Protocol 6: Confirmation Menu
Present structured options:
- **[A] Accept** - Confirm and advance to next step
- **[P] Pivot** - Modify type/metadata choices
- **[C] Clarify** - Ask questions about classification
- **[P] Pivot** - Modify sidecar/metadata choices
- **[C] Clarify** - Ask questions about sidecar decision
---
# CONTEXT BOUNDARIES
## In Scope
- Agent type classification
- Sidecar decision (hasSidecar: true/false)
- All 6 metadata properties
- Documentation to plan file
- Type selection guidance with examples
- Sidecar decision guidance with examples
## Out of Scope (Future Steps)
- Persona/character development (Step 3)
- Command structure design (Step 4)
- Agent naming/branding refinement (Step 5)
- Implementation/build (Step 6)
- Validation/testing (Step 7)
- Persona/character development (Step 4)
- Command structure design (Step 5)
- Agent naming/branding refinement (Step 6)
- Implementation/build (Step 7)
- Validation/testing (Step 8)
## Red Flags to Address
- User wants complex agent but selects "Simple" type
- Module classification without workflow management needs
- User wants complex memory but selects hasSidecar: false
- Unclear about what "memory across sessions" means
- Missing or unclear metadata properties
- Module path format confusion
---
## MANDATORY SEQUENCE
# MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
## 1. Load Documentation
Read and internalize:
- `{agentTypesDoc}` - Classification framework
- `{agentTypesDoc}` - Sidecar decision framework
- `{agentMetadata}` - Property definitions
- Keep examples accessible for reference
## 2. Purpose Discovery Conversation
## 2. Sidecar Decision Conversation
Engage user with questions in `{communication_language}`:
- "What is the primary function this agent will perform?"
- "How complex are the tasks this agent will handle?"
- "Will this agent need to manage workflows or other agents?"
- "What specific domains or expertise areas are involved?"
- "Should your agent remember things between sessions?"
- "What should it remember? User preferences? Project state? Progress over time?"
- "Or is each interaction independent and fresh?"
Listen for natural language cues about scope and complexity.
Listen for natural language cues about memory needs.
## 3. Agent Type Determination
Based on discovery, propose classification:
- Present recommended type with reasoning
## 3. Sidecar Determination
Based on discovery, propose decision:
- Present recommended hasSidecar value with reasoning
- Show relevant example if helpful
- Confirm classification matches user intent
- Confirm decision matches user intent
- Allow pivoting if user vision evolves
**Conversation Template:**
```
Based on our discussion, I recommend classifying this as a [TYPE] agent because:
Based on our discussion, I recommend hasSidecar: [true/false] because:
[reasoning from discovery]
[If helpful: "For reference, here's a similar [TYPE] agent:"]
[Show relevant example path: simpleExample/expertExample/moduleExample]
[If helpful: "For reference, here's a similar agent:"]
[Show relevant example path: noSidecarExample/withSidecarExample]
Does this classification feel right to you?
Does this feel right to you?
```
## 4. Define All Metadata Properties
@@ -196,10 +208,10 @@ Work through each property systematically:
- Guide user through structure if unfamiliar
**4f. Sidecar Configuration**
- Boolean: manages external workflows?
- Typically false for Simple/Expert agents
- True for Module agents that deploy workflows
- Confirm based on user's integration needs
- Boolean: does agent need memory?
- Most personality-driven agents don't need it
- Most relationship/coaching/tracking agents do need it
- Confirm based on user's memory needs
**Conversation Template:**
```
@@ -221,10 +233,10 @@ Ready to confirm, or should we adjust any properties?
Write to `{agentPlan}`:
```yaml
# Agent Type & Metadata
agent_type: [Simple|Expert|Module]
classification_rationale: |
[Clear explanation of why this type matches user's articulated needs]
# Agent Sidecar Decision & Metadata
hasSidecar: [true|false]
sidecar_rationale: |
[Clear explanation of why this agent does or does not need memory based on user's stated needs]
metadata:
id: [technical-identifier]
@@ -234,12 +246,12 @@ metadata:
module: [project:type:name]
hasSidecar: [true|false]
# Type Classification Notes
type_decision_date: [YYYY-MM-DD]
type_confidence: [High/Medium/Low]
considered_alternatives: |
- [Alternative type]: [reason not chosen]
- [Alternative type]: [reason not chosen]
# Sidecar Decision Notes
sidecar_decision_date: [YYYY-MM-DD]
sidecar_confidence: [High/Medium/Low]
memory_needs_identified: |
- [Specific memory needs if hasSidecar: true]
- [Or: N/A - stateless interactions]
```
### 6. Present MENU OPTIONS
@@ -262,14 +274,14 @@ Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Cont
## CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [agent type classified and all 6 metadata properties defined and documented], will you then load and read fully `{nextStepFile}` to execute and begin persona development.
ONLY WHEN [C continue option] is selected and [hasSidecar decision made and all 6 metadata properties defined and documented], will you then load and read fully `{nextStepFile}` to execute and begin persona development.
---
# SYSTEM SUCCESS/FAILURE METRICS
## Success Indicators
- Type classification clearly justified
- Sidecar decision clearly justified
- All metadata properties populated correctly
- YAML structure matches specification exactly
- User confirms understanding and acceptance
@@ -278,16 +290,16 @@ ONLY WHEN [C continue option] is selected and [agent type classified and all 6 m
## Failure Indicators
- Missing or undefined metadata properties
- YAML structure malformed
- User confusion about type classification
- User confusion about sidecar decision
- Inadequate documentation to plan file
- Proceeding without user confirmation
## Recovery Mode
If user struggles with classification:
If user struggles with sidecar decision:
- Show concrete examples from each type
- Compare/contrast types with their use case
- Ask targeted questions about complexity/scope
- Offer type recommendation with clear reasoning
- Compare/contrast with their use case
- Ask targeted questions about memory needs
- Offer recommendation with clear reasoning
Recover metadata definition issues by:
- Showing property format examples
_bmad/bmb/workflows/agent/steps-c/step-04-persona.md
+2 -2
View File
@@ -10,8 +10,8 @@ principlesCrafting: ../data/principles-crafting.md
communicationPresets: ../data/communication-presets.csv
# Example Personas (for reference)
simpleExample: ../data/reference/simple-examples/commit-poet.agent.yaml
expertExample: ../data/reference/expert-examples/journal-keeper/journal-keeper.agent.yaml
simpleExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
expertExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
_bmad/bmb/workflows/agent/steps-c/step-05-commands-menu.md
+2 -2
View File
@@ -8,8 +8,8 @@ agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
agentMenuPatterns: ../data/agent-menu-patterns.md
# Example Menus (for reference)
simpleExample: ../data/reference/simple-examples/commit-poet.agent.yaml
expertExample: ../data/reference/expert-examples/journal-keeper/journal-keeper.agent.yaml
simpleExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
expertExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
_bmad/bmb/workflows/agent/steps-c/step-06-activation.md
+80 -82
View File
@@ -6,13 +6,12 @@ description: 'Plan activation behavior and route to build'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
criticalActions: ../data/critical-actions.md
# Build Step Routes (determined by agent type)
simpleBuild: './step-07a-build-simple.md'
expertBuild: './step-07b-build-expert.md'
moduleBuild: './step-07c-build-module.md'
# Build Step Route (determined by hasSidecar)
agentBuild: './step-07-build-agent.md'
# Example critical_actions (for reference)
expertExample: ../data/reference/expert-examples/journal-keeper/journal-keeper.agent.yaml
withSidecarExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
withoutSidecarExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
@@ -20,7 +19,8 @@ partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# STEP GOAL
Define activation behavior through critical_actions and route to the appropriate build step based on agent complexity.
Define activation behavior through critical_actions and confirm routing to the build step based on hasSidecar decision.
# MANDATORY EXECUTION RULES
@@ -29,9 +29,9 @@ Define activation behavior through critical_actions and route to the appropriate
- Read agentPlan to access all accumulated metadata
- These are non-negotiable prerequisites
2. **MUST Determine Route Before Activation Discussion**
- Check `module` and `hasSidecar` from plan metadata
- Determine destination build step FIRST
2. **MUST Confirm hasSidecar Decision**
- Check `hasSidecar` from plan metadata (decided in Step 3)
- This determines the build approach
- Inform user of routing decision
3. **MUST Document Activation Decision**
@@ -39,14 +39,11 @@ Define activation behavior through critical_actions and route to the appropriate
- OR document deliberate omission with rationale
- No middle ground - commit to one path
4. **MUST Follow Routing Logic Exactly**
4. **MUST Follow Simple Routing Logic**
```yaml
# Route determination based on module and hasSidecar
# Module agents: any module value other than "stand-alone"
module ≠ "stand-alone" → step-07c-build-module.md
# Stand-alone agents: determined by hasSidecar
module = "stand-alone" + hasSidecar: true → step-07b-build-expert.md
module = "stand-alone" + hasSidecar: false → step-07a-build-simple.md
# Route determination based on hasSidecar only
hasSidecar: false → Agent without sidecar (single YAML file)
hasSidecar: true → Agent with sidecar (YAML + sidecar folder)
```
5. **NEVER Skip Documentation**
@@ -62,10 +59,9 @@ Execute BEFORE engaging user:
1. Load criticalActions.md
2. Load agentPlan-{agent_name}.md
3. Extract routing metadata:
- hasSidecar (boolean)
- module (string)
- agentType (if defined)
4. Determine destination build step
- hasSidecar (boolean) - decided in Step 3
- All other metadata from prior steps
4. Confirm build approach
## Protocol 2: Routing Disclosure
Inform user immediately of determined route:
@@ -73,9 +69,8 @@ Inform user immediately of determined route:
```
"Based on your agent configuration:
- hasSidecar: {hasSidecar}
- module: {module}
→ Routing to: {destinationStep}
→ Building: Agent {WITH|WITHOUT} sidecar
Now let's plan your activation behavior..."
```
@@ -106,10 +101,10 @@ activation:
hasCriticalActions: true/false
rationale: "Explanation of why or why not"
criticalActions: [] # Only if hasCriticalActions: true
routing:
destinationBuild: "step-06-{X}.md"
buildApproach: "Agent {with|without} sidecar"
hasSidecar: {boolean}
module: "{module}"
```
# CONTEXT BOUNDARIES
@@ -117,21 +112,22 @@ routing:
## In Scope
- Planning activation behavior for the agent
- Defining critical_actions array
- Routing to appropriate build step
- Confirming routing to build step
- Documenting activation decisions
## Out of Scope
- Writing actual activation code (build step)
- Designing sidecar workflows (build step)
- Changing core agent metadata (locked after step 04)
- Changing core agent metadata (locked after Step 4)
- Implementing commands (build step)
## Routing Boundaries
- Simple agents: No sidecar, straightforward activation
- Expert agents: Sidecar + stand-alone module
- Module agents: Sidecar + parent module integration
- **Agent WITHOUT sidecar**: Single YAML file, no persistent memory
- **Agent WITH sidecar**: YAML file + sidecar folder with persistent memory
## MANDATORY SEQUENCE
---
# MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
@@ -142,14 +138,26 @@ cat {criticalActions}
cat {agentPlan}
```
## 2. Discuss Activation Needs
## 2. Confirm Routing Decision
Verify hasSidecar decision from Step 3:
```
"Confirming your agent configuration from Step 3:
- hasSidecar: {value from plan}
- This means: {Agent will|will not} remember things between sessions
- Build approach: {Single YAML file|YAML + sidecar folder}
Is this still correct?"
```
## 3. Discuss Activation Needs
Ask user:
- "Should your agent be able to take autonomous actions?"
- "Are there specific workflows it should trigger?"
- "Should it run as a background process or scheduled task?"
- "Or will it primarily respond to direct prompts?"
## 3. Define critical_actions OR Explicitly Omit
## 4. Define critical_actions OR Explicitly Omit
**If defining:**
- Reference criticalActions.md patterns
@@ -157,29 +165,27 @@ Ask user:
- Each action should be clear and scoped
- Document rationale for each
**For agents WITH sidecar, critical_actions MUST include:**
```
- "Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/memories.md"
- "Load COMPLETE file {project-root}/_bmad/_memory/{sidecar-folder}/instructions.md"
- "ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/ - private space"
```
Plus any additional activation behaviors the agent needs.
**For agents WITHOUT sidecar, critical_actions are OPTIONAL and can include:**
```
- "Give user an inspirational quote before showing menu"
- "Fetch latest data from {project-root}/finances/ before displaying menu"
- "Display a quick status summary on activation"
```
Agents without sidecar omit critical_actions entirely if no activation behavior is needed.
**If omitting:**
- State clearly: "This agent will not have critical_actions"
- Explain why: "This agent is a responsive assistant that operates under direct user guidance"
- Document the rationale
## 4. Route to Build Step
Determine destination:
```yaml
# Check plan metadata
hasSidecar: {value from step 04}
module: "{value from step 04}"
# Route logic
if hasSidecar == false:
destination = simpleBuild
elif hasSidecar == true and module == "stand-alone":
destination = expertBuild
else: # hasSidecar == true and module != "stand-alone"
destination = moduleBuild
```
## 5. Document to Plan
Update agentPlan with:
@@ -187,21 +193,17 @@ Update agentPlan with:
```yaml
---
activation:
hasCriticalActions: true
rationale: "Agent needs to autonomously trigger workflows for task automation"
hasCriticalActions: {true/false}
rationale: "Agent needs to autonomously trigger workflows for task automation" OR "Agent operates under direct user guidance"
criticalActions:
- name: "start-workflow"
description: "Initiate a predefined workflow for task execution"
- name: "schedule-task"
description: "Schedule tasks for future execution"
- name: "sync-data"
description: "Synchronize data with external systems"
# ... additional actions if needed
routing:
destinationBuild: "step-06-build-expert.md"
hasSidecar: true
module: "stand-alone"
rationale: "Agent requires sidecar workflows for autonomous operation"
buildApproach: "Agent {with|without} sidecar"
hasSidecar: {true/false}
rationale: "Agent {needs|does not need} persistent memory across sessions"
---
```
@@ -213,7 +215,7 @@ Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Cont
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Save content to {agentPlan}, update frontmatter, determine appropriate build step based on hasSidecar and module values, then only then load, read entire file, then execute {simpleBuild} or {expertBuild} or {moduleBuild} as determined
- IF C: Save content to {agentPlan}, update frontmatter, then only then load, read entire file, then execute {agentBuild}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#6-present-menu-options)
#### EXECUTION RULES:
@@ -225,14 +227,13 @@ Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Cont
## CRITICAL STEP COMPLETION NOTE
This is the **ROUTING HUB** of agent creation. ONLY WHEN [C continue option] is selected and [routing decision determined with activation needs documented], will you then determine the appropriate build step based on hasSidecar/module values and load and read fully that build step file to execute.
This is the **final planning step** before building. ONLY WHEN [C continue option] is selected and [activation needs documented], will you then load and read fully `{agentBuild}` to execute and build the agent.
Routing logic:
- hasSidecar: false → step-06-build-simple.md
- hasSidecar: true + module: "stand-alone" → step-06-build-expert.md
- hasSidecar: true + module: ≠ "stand-alone" → step-06-build-module.md
- hasSidecar: false → Agent WITHOUT sidecar (single YAML)
- hasSidecar: true → Agent WITH sidecar (YAML + sidecar folder)
You cannot proceed to build without completing routing.
You cannot proceed to build without completing activation planning.
---
@@ -241,39 +242,36 @@ You cannot proceed to build without completing routing.
✅ **COMPLETION CRITERIA:**
- [ ] criticalActions.md loaded and understood
- [ ] agentPlan loaded with all prior metadata
- [ ] Routing decision determined and communicated
- [ ] Routing decision confirmed (hasSidecar from Step 3)
- [ ] Activation needs discussed with user
- [ ] critical_actions defined OR explicitly omitted with rationale
- [ ] Plan updated with activation and routing metadata
- [ ] User confirms routing to appropriate build step
- [ ] User confirms ready to build
✅ **SUCCESS INDICATORS:**
- Clear activation decision documented
- Route to build step is unambiguous
- User understands why they're going to {simple|expert|module} build
- Route to build is unambiguous
- User understands the build approach
- Plan file reflects complete activation configuration
❌ **FAILURE MODES:**
- Attempting to define critical_actions without reading reference
- Routing decision not documented in plan
- User doesn't understand which build step comes next
- User doesn't understand the build approach
- Ambiguous activation configuration (neither defined nor omitted)
- Skipping routing discussion entirely
- Skipping activation discussion entirely
⚠️ **RECOVERY PATHS:**
⚠️ **RECOVERY PATHS**
If activation planning goes wrong:
1. **Can't decide on activation?**
- Default: Omit critical_actions
- Route to simpleBuild
- Can add later via edit-agent workflow
2. **Uncertain about routing?**
- Check hasSidecar value
- Check module value
- Apply routing logic strictly
3. **User wants to change route?**
- Adjust hasSidecar or module values
- Re-run routing logic
2. **User wants to change hasSidecar?**
- Return to Step 3 to revise decision
- Update plan accordingly
3. **Uncertain about routing?**
- Check hasSidecar value
- Apply simple routing logic
_bmad/bmb/workflows/agent/steps-c/step-07-build-agent.md
+315
View File
@@ -0,0 +1,315 @@
---
name: 'step-07-build-agent'
description: 'Generate agent YAML from plan (with or without sidecar)'
# File References
nextStepFile: './step-08-celebrate.md'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
# Output paths (determined by hasSidecar)
agentBuildOutput: '{bmb_creations_output_folder}/{agent-name}/'
agentYamlOutput: '{bmb_creations_output_folder}/{agent-name}/{agent-name}.agent.yaml'
agentYamlOutputNoSidecar: '{bmb_creations_output_folder}/{agent-name}.agent.yaml'
sidecarOutput: '{bmb_creations_output_folder}/{agent-name}/{agent-name}-sidecar/'
# Template and Architecture
agentTemplate: ../templates/agent-template.md
agentArch: ../data/agent-architecture.md
agentCompilation: ../data/agent-compilation.md
criticalActions: ../data/critical-actions.md
# Reference examples
noSidecarExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
withSidecarExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# STEP GOAL
Assemble the agent plan content into a complete agent YAML file. The build approach (with or without sidecar) is determined by the `hasSidecar` decision made in Step 3.
---
# MANDATORY EXECUTION RULES
1. **DETERMINE BUILD APPROACH FIRST**: Check `hasSidecar` from agentPlan before starting
2. **TEMPLATE COMPLIANCE**: Follow agent-template.md structure exactly
3. **YAML VALIDATION**: Ensure valid YAML syntax with proper indentation (2-space)
4. **EXISTING CHECK**: If output file exists, ask user before overwriting
5. **NO DRIFT**: Use ONLY content from agentPlan - no additions or interpretations
6. **SIDECAR REQUIREMENT**: If hasSidecar=true, MUST create sidecar folder structure
---
# EXECUTION PROTOCOLS
## Phase 1: Load Architecture and Templates
1. Read `agentTemplate` - defines YAML structure for agents
2. Read `agentArch` - architecture requirements for agents
3. Read `agentCompilation` - assembly rules for YAML generation
4. Read `criticalActions` - validation requirements for critical_actions
## Phase 2: Load Agent Plan
1. Read `agentPlan` containing all collected content from Steps 2-5
2. Verify plan contains:
- hasSidecar decision (true/false)
- Persona content
- Commands structure
- All metadata fields
- Activation decisions (critical_actions)
## Phase 3: Determine Build Approach
Check `hasSidecar` from plan:
```yaml
hasSidecar: false
→ Build: Agent WITHOUT sidecar
→ Output: Single YAML file at {agentYamlOutputNoSidecar}
→ Structure: Everything in one file (~250 lines max)
hasSidecar: true
→ Build: Agent WITH sidecar
→ Output: YAML + sidecar folder structure
→ Structure: YAML file + {agent-name}-sidecar/ folder
```
**Inform user of build approach:**
```
"Building: Agent {WITH|WITHOUT} sidecar
hasSidecar: {true/false}
Output: {output path description}"
```
## Phase 4: Assemble Agent YAML
### For Agents WITHOUT Sidecar (hasSidecar: false)
**Structure:**
```yaml
name: '{agent-name}'
description: '{short-description}'
author:
name: '{author}'
created: '{date}'
persona: |
{multi-line persona content from plan}
system-context: |
{expanded context from plan}
capabilities:
- {capability from plan}
- {capability from plan}
# ... all capabilities
commands:
- name: '{command-name}'
description: '{what command does}'
trigger: '{menu trigger}'
steps:
- {step 1}
- {step 2}
# ... all commands from plan
configuration:
temperature: {temperature}
max-tokens: {max-tokens}
response-format: {format}
# ... other configuration from plan
metadata:
hasSidecar: false
agent-type: 'agent'
```
**Output:** Single YAML file at `{agentYamlOutputNoSidecar}`
### For Agents WITH Sidecar (hasSidecar: true)
**Structure:**
```yaml
name: '{agent-name}'
description: '{short-description}'
author:
name: '{author}'
created: '{date}'
persona: |
{multi-line persona content from plan}
system-context: |
{expanded context from plan}
capabilities:
- {capability from plan}
- {capability from plan}
# ... all capabilities
critical-actions:
- name: '{action-name}'
description: '{what it does}'
invocation: '{when/how to invoke}'
implementation: |
{multi-line implementation}
output: '{expected-output}'
sidecar-folder: '{sidecar-folder-name}'
sidecar-files:
- '{project-root}/_bmad/_memory/{sidecar-folder}/{file1}.md'
- '{project-root}/_bmad/_memory/{sidecar-folder}/{file2}.md'
# ... all critical actions referencing sidecar structure
commands:
- name: '{command-name}'
description: '{what command does}'
trigger: '{menu trigger}'
steps:
- {step 1}
- {step 2}
# ... all commands from plan
configuration:
temperature: {temperature}
max-tokens: {max-tokens}
response-format: {format}
# ... other configuration from plan
metadata:
sidecar-folder: '{sidecar-folder-name}'
sidecar-path: '{project-root}/_bmad/_memory/{sidecar-folder}/'
hasSidecar: true
agent-type: 'agent'
memory-type: 'persistent'
```
**Output:** YAML file at `{agentYamlOutput}` + sidecar folder structure
### Phase 5: Create Sidecar Structure (IF hasSidecar: true)
Skip this phase if hasSidecar: false
1. **Create Sidecar Directory**:
```bash
mkdir -p {sidecarOutput}
```
2. **Create Starter Files** (if specified in critical_actions):
```bash
touch {sidecarOutput}/memories.md
touch {sidecarOutput}/instructions.md
# ... additional files from critical_actions
```
3. **Add README to Sidecar**:
```markdown
# {sidecar-folder} Sidecar
This folder stores persistent memory for the **{agent-name}** agent.
## Purpose
{purpose from critical_actions}
## Files
- memories.md: User profile, session history, patterns
- instructions.md: Protocols, boundaries, startup behavior
- {additional files}
## Runtime Access
After BMAD installation, this folder will be accessible at:
`{project-root}/_bmad/_memory/{sidecar-folder}/{filename}.md`
```
### Phase 6: Write Agent YAML
**If hasSidecar: false:**
1. Write YAML to `{agentYamlOutputNoSidecar}`
2. Confirm write success
3. Display file location to user
**If hasSidecar: true:**
1. Create directory: `mkdir -p {agentBuildOutput}`
2. Write YAML to `{agentYamlOutput}`
3. Confirm write success
4. Display file location to user
## Phase 7: Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Write agent YAML to appropriate output path (with or without sidecar), update frontmatter, then only then load, read entire file, then execute {nextStepFile}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#7-present-menu-options)
#### EXECUTION RULES:
- ALWAYS halt and wait for user input after presenting menu
- ONLY proceed to next step when user selects 'C'
- After other menu items execution, return to this menu
- User can chat or ask questions - always respond and then end with display again of the menu options
---
# CONTEXT BOUNDARIES
**INCLUDE:**
- Template structure exactly as provided
- All agent metadata from agentPlan
- Persona, commands, and rules from plan
- Configuration options specified
- Sidecar structure if hasSidecar: true
**EXCLUDE:**
- Any content not in agentPlan
- Sidecar references if hasSidecar: false
- Template placeholders (replace with actual content)
- Comments or notes in final YAML
---
# CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [complete YAML generated and written to output], will you then load and read fully `{nextStepFile}` to execute and celebrate completion.
**This step produces:**
- **If hasSidecar: false**: Single agent YAML file
- **If hasSidecar: true**: Agent YAML file + sidecar folder structure
Both must exist (if applicable) before proceeding to validation.
---
# SUCCESS METRICS
✅ **SUCCESS looks like:**
- Agent YAML file exists at specified output path
- YAML is syntactically valid and well-formed
- All template fields populated with plan content
- Structure matches agent architecture
- If hasSidecar: true, sidecar folder created with starter files
- User has selected continue to proceed
❌ **FAILURE looks like:**
- Template or architecture files not found
- Agent plan missing required sections
- YAML syntax errors in output
- Content not properly mapped to template
- File write operation fails
- hasSidecar: true but sidecar folder not created
---
# TRANSITION CRITERIA
**Ready for Step 8 when:**
- Agent YAML successfully created (with or without sidecar as specified)
- User selects continue
- All build artifacts confirmed written
_bmad/bmb/workflows/agent/steps-c/step-07a-build-simple.md
-187
View File
@@ -1,187 +0,0 @@
---
name: 'step-07a-build-simple'
description: 'Generate Simple agent YAML from plan'
# File References
nextStepFile: './step-08-celebrate.md'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
agentBuildOutput: '{bmb_creations_output_folder}/{agent-name}.agent.yaml'
# Template and Architecture
simpleTemplate: ../templates/simple-agent.template.md
simpleArch: ../data/simple-agent-architecture.md
agentCompilation: ../data/agent-compilation.md
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# STEP GOAL
Assemble the agent plan content into a Simple agent YAML configuration using the template, producing a complete agent definition ready for validation.
## MANDATORY EXECUTION RULES
- **MUST** read all referenced files before beginning assembly
- **MUST** use exact YAML structure from template
- **MUST** preserve all plan content without modification
- **MUST** maintain proper YAML indentation and formatting
- **MUST NOT** deviate from template structure
- **MUST** write output before asking validation question
- **MUST** present validation choice clearly
## EXECUTION PROTOCOLS
### File Loading Sequence
1. Read `simpleTemplate` - provides the YAML structure
2. Read `simpleArch` - defines Simple agent architecture rules
3. Read `agentCompilation` - provides assembly guidelines
4. Read `agentPlan` - contains structured content from steps 2-5
### YAML Assembly Process
1. Parse template structure
2. Extract content sections from agentPlan YAML
3. Map plan content to template fields
4. Validate YAML syntax before writing
5. Write complete agent YAML to output path
## CONTEXT BOUNDARIES
**INCLUDE:**
- Template structure exactly as provided
- All agent metadata from agentPlan
- Persona, commands, and rules from plan
- Configuration options specified
**EXCLUDE:**
- Any content not in agentPlan
- Sidecar file references (Simple agents don't use them)
- Template placeholders (replace with actual content)
- Comments or notes in final YAML
## MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
### 1. Load Template and Architecture Files
Read the following files in order:
- `simpleTemplate` - YAML structure template
- `simpleArch` - Simple agent architecture definition
- `agentCompilation` - Assembly instructions
**Verify:** All files loaded successfully.
### 2. Load Agent Plan
Read `agentPlan` which contains structured YAML from steps 2-5:
- Step 2: Discovery findings
- Step 3: Persona development
- Step 4: Command structure
- Step 5: Agent naming
**Verify:** Plan contains all required sections.
### 3. Assemble YAML Using Template
Execute the following assembly process:
1. **Parse Template Structure**
- Identify all YAML fields
- Note required vs optional fields
- Map field types and formats
2. **Extract Plan Content**
- Read agent metadata
- Extract persona definition
- Retrieve command specifications
- Gather rules and constraints
3. **Map Content to Template**
- Replace template placeholders with plan content
- Maintain exact YAML structure
- Preserve indentation and formatting
- Validate field types and values
4. **Validate YAML Syntax**
- Check proper indentation
- Verify quote usage
- Ensure list formatting
- Confirm no syntax errors
**Verify:** YAML is valid, complete, and follows template structure.
### 4. Write Agent Build Output
Write the assembled YAML to `agentBuildOutput`:
- Use exact output path from variable
- Include all content without truncation
- Maintain YAML formatting
- Confirm write operation succeeded
**Verify:** File written successfully and contains complete YAML.
### 5. Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Write agent YAML to {agentBuildOutput}/{agent-name}.agent.yaml (or appropriate output path), update frontmatter, then only then load, read entire file, then execute {nextStepFile}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#5-present-menu-options)
#### EXECUTION RULES:
- ALWAYS halt and wait for user input after presenting menu
- ONLY proceed to next step when user selects 'C'
- After other menu items execution, return to this menu
- User can chat or ask questions - always respond and then end with display again of the menu options
### 6. Route Based on User Choice
**If user chooses "one-at-a-time":**
- Proceed to `nextStepFile` (step-08-celebrate.md)
- Continue through each validation step sequentially
- Allow review between each validation
**If user chooses "YOLO":**
- Run all validation steps (7A through 7F) consecutively
- Do not pause between validations
- After all validations complete, proceed to Step 8
- Present summary of all validation results
## CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [complete YAML generated and written to output], will you then load and read fully `{nextStepFile}` to execute and celebrate completion.
## SUCCESS METRICS
**SUCCESS looks like:**
- Agent YAML file exists at specified output path
- YAML is syntactically valid and well-formed
- All template fields populated with plan content
- Structure matches Simple agent architecture
- User has selected validation approach
- Clear next step identified
**FAILURE looks like:**
- Template or architecture files not found
- Agent plan missing required sections
- YAML syntax errors in output
- Content not properly mapped to template
- File write operation fails
- User selection unclear
## TRANSITION CRITERIA
**Ready for Step 7A when:**
- Simple agent YAML successfully created
- User chooses "one-at-a-time" validation
**Ready for Step 8 when:**
- Simple agent YAML successfully created
- User chooses "YOLO" validation
- All validations (7A-7F) completed consecutively
_bmad/bmb/workflows/agent/steps-c/step-07b-build-expert.md
-201
View File
@@ -1,201 +0,0 @@
---
name: 'step-06-build-expert'
description: 'Generate Expert agent YAML with sidecar from plan'
# File References
nextStepFile: './step-08-celebrate.md'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
agentBuildOutput: '{bmb_creations_output_folder}/{agent-name}/'
agentYamlOutput: '{bmb_creations_output_folder}/{agent-name}/{agent-name}.agent.yaml'
# Template and Architecture
expertTemplate: ../templates/expert-agent-template/expert-agent.template.md
expertArch: ../data/expert-agent-architecture.md
agentCompilation: ../data/agent-compilation.md
criticalActions: ../data/critical-actions.md
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# STEP GOAL
Assemble the agent plan content into a complete Expert agent YAML file with sidecar folder structure. Expert agents require persistent memory storage, so the build creates a sidecar folder next to the agent.yaml (which gets installed to `_bmad/_memory/` during BMAD installation).
## MANDATORY EXECUTION RULES
1. **EXPERT AGENT = SIDECAR REQUIRED**: Every Expert agent MUST have a sidecar folder created next to agent.yaml (build location), which will be installed to `_bmad/_memory/` during BMAD installation
2. **CRITICAL_ACTIONS FORMAT**: All critical_actions MUST use `{project-root}/_bmad/_memory/{sidecar-folder}/` for file operations (runtime path)
3. **TEMPLATE COMPLIANCE**: Follow expert-agent-template.md structure exactly
4. **YAML VALIDATION**: Ensure valid YAML syntax with proper indentation (2-space)
5. **EXISTING CHECK**: If agentYamlOutput exists, ask user before overwriting
6. **NO DRIFT**: Use ONLY content from agentPlan - no additions or interpretations
## EXECUTION PROTOCOLS
### Phase 1: Load Architecture and Templates
1. Read `expertTemplate` - defines YAML structure for Expert agents
2. Read `expertArch` - architecture requirements for Expert-level agents
3. Read `agentCompilation` - assembly rules for YAML generation
4. Read `criticalActions` - validation requirements for critical_actions
### Phase 2: Load Agent Plan
1. Read `agentPlan` containing all collected content from Steps 1-5
2. Verify plan contains:
- Agent type: "expert"
- Sidecar folder name
- Persona content
- Commands structure
- Critical actions (if applicable)
### Phase 3: Assemble Expert YAML
Using expertTemplate as structure:
```yaml
name: '{agent-name}'
description: '{short-description}'
author:
name: '{author}'
created: '{date}'
persona: |
{multi-line persona content from plan}
system-context: |
{expanded context from plan}
capabilities:
- {capability from plan}
- {capability from plan}
# ... all capabilities
critical-actions:
- name: '{action-name}'
description: '{what it does}'
invocation: '{when/how to invoke}'
implementation: |
{multi-line implementation}
output: '{expected-output}'
sidecar-folder: '{sidecar-folder-name}'
sidecar-files:
- '{project-root}/_bmad/_memory/{sidecar-folder}/{file1}.md'
- '{project-root}/_bmad/_memory/{sidecar-folder}/{file2}.md'
# ... all critical actions referencing sidecar structure
commands:
- name: '{command-name}'
description: '{what command does}'
steps:
- {step 1}
- {step 2}
# ... all commands from plan
configuration:
temperature: {temperature}
max-tokens: {max-tokens}
response-format: {format}
# ... other configuration from plan
metadata:
sidecar-folder: '{sidecar-folder-name}'
sidecar-path: '{project-root}/_bmad/_memory/{sidecar-folder}/'
agent-type: 'expert'
memory-type: 'persistent'
```
### Phase 4: Create Sidecar Structure
1. **Create Sidecar Directory** (NEXT TO agent.yaml):
- Path: `{agentBuildOutput}/{agent-name}-sidecar/`
- Use `mkdir -p` to create full path
- Note: This folder gets installed to `_bmad/_memory/` during BMAD installation
2. **Create Starter Files** (if specified in critical_actions):
```bash
touch {agentBuildOutput}/{agent-name}-sidecar/{file1}.md
touch {agentBuildOutput}/{agent-name}-sidecar/{file2}.md
```
3. **Add README to Sidecar**:
```markdown
# {sidecar-folder} Sidecar
This folder stores persistent memory for the **{agent-name}** Expert agent.
## Purpose
{purpose from critical_actions}
## Files
- {file1}.md: {description}
- {file2}.md: {description}
## Runtime Access
After BMAD installation, this folder will be accessible at:
`{project-root}/_bmad/_memory/{sidecar-folder}/{filename}.md`
```
### Phase 5: Write Agent YAML
1. Create `agentBuildOutput` directory: `mkdir -p {agentBuildOutput}`
2. Write YAML to `agentYamlOutput`
3. Confirm write success
4. Display file location to user
### Phase 6: Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Write agent YAML to {agentBuildOutput}/{agent-name}/{agent-name}.agent.yaml (or appropriate output path), update frontmatter, then only then load, read entire file, then execute {nextStepFile}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#phase-6-present-menu-options)
#### EXECUTION RULES:
- ALWAYS halt and wait for user input after presenting menu
- ONLY proceed to next step when user selects 'C'
- After other menu items execution, return to this menu
- User can chat or ask questions - always respond and then end with display again of the menu options
## CONTEXT BOUNDARIES
- **USE ONLY**: Content from agentPlan, expertTemplate, expertArch, agentCompilation, criticalActions
- **DO NOT ADD**: New capabilities, commands, or actions not in plan
- **DO NOT INTERPRET**: Use exact language from plan
- **DO NOT SKIP**: Any field in expertTemplate structure
- **CRITICAL**: Expert agents MUST have sidecar-folder metadata
## CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [complete YAML generated and written to output], will you then load and read fully `{nextStepFile}` to execute and celebrate completion.
This step produces TWO artifacts:
1. **Agent YAML**: Complete expert agent definition at `{agentYamlOutput}`
2. **Sidecar Structure**: Folder and files at `{agentBuildOutput}/{agent-name}-sidecar/` (build location, installs to `_bmad/_memory/` during BMAD installation)
Both must exist before proceeding to validation.
## SUCCESS METRICS
✅ Agent YAML file created at expected location
✅ Valid YAML syntax (no parse errors)
✅ All template fields populated
✅ Sidecar folder created at `{agentBuildOutput}/{agent-name}-sidecar/` (build location)
✅ Sidecar folder contains starter files from critical_actions
✅ critical_actions reference `{project-root}/_bmad/_memory/{sidecar-folder}/` paths
✅ metadata.sidecar-folder populated
✅ metadata.agent-type = "expert"
✅ User validation choice received (one-at-a-time or YOLO)
## FAILURE MODES
❌ Missing required template fields
❌ Invalid YAML syntax
❌ Sidecar folder creation failed
❌ critical_actions missing sidecar-folder references
❌ agentPlan missing expert-specific content (sidecar-folder name)
❌ File write permission errors
_bmad/bmb/workflows/agent/steps-c/step-07c-build-module.md
-258
View File
@@ -1,258 +0,0 @@
---
name: 'step-06-build-module'
description: 'Generate Module agent YAML from plan'
# File References
nextStepFile: './step-08-celebrate.md'
agentPlan: '{bmb_creations_output_folder}/agent-plan-{agent_name}.md'
agentBuildOutput: '{bmb_creations_output_folder}/{agent-name}/'
agentYamlOutput: '{bmb_creations_output_folder}/{agent-name}/{agent-name}.agent.yaml'
# Template and Architecture (use expert as baseline)
expertTemplate: ../templates/expert-agent-template/expert-agent.template.md
expertArch: ../data/expert-agent-architecture.md
agentCompilation: ../data/agent-compilation.md
criticalActions: ../data/critical-actions.md
# Task References
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# STEP GOAL
Assemble the Module agent YAML file from the approved plan, using the expert agent template as the baseline architecture and adding module-specific workflow integration paths and sidecar configuration.
# MANDATORY EXECUTION RULES
1. **TEMPLATE BASELINE**: Module agents MUST use the expert agent template as their structural foundation - do not create custom templates
2. **PLAN ADHERENCE**: Extract content from agentPlan exactly as written - no enhancement, interpretation, or extrapolation
3. **MODULE SPECIFICITY**: Module agents require workflow integration paths and may need sidecar configuration for multi-workflow modules
4. **OUTPUT VALIDATION**: YAML must be valid, complete, and ready for immediate deployment
5. **LANGUAGE PRESERVATION**: Maintain any language choice configured in the plan throughout the YAML
# EXECUTION PROTOCOLS
## PREPARATION PHASE
### 1. Load Expert Template Baseline
```
Read: expertTemplate
Read: expertArch
Read: agentCompilation
Read: criticalActions
```
**Purpose**: Understand the expert agent structure that serves as the Module agent baseline
**Validation**: Confirm expert template has all required sections (name, description, persona, instructions, tools, skills, etc.)
### 2. Load Agent Plan
```
Read: agentPlan (using dynamic path)
```
**Validation**: Plan contains all mandatory sections:
- Agent identity (name, description)
- Persona profile
- Command structure
- Critical actions
- Workflow integrations (module-specific)
- Language choice (if configured)
### 3. Verify Output Directory
```
Bash: mkdir -p {agentBuildOutput}
```
**Purpose**: Ensure output directory exists for the module agent
## ASSEMBLY PHASE
### 4. Assemble Module Agent YAML
**FROM PLAN TO YAML MAPPING:**
| Plan Section | YAML Field | Notes |
|--------------|------------|-------|
| Agent Name | `name` | Plan → YAML |
| Description | `description` | Plan → YAML |
| Persona | `persona` | Plan → YAML |
| Instructions | `instructions` | Plan → YAML (verbatim) |
| Commands | `commands` | Plan → YAML (with handlers) |
| Critical Actions | `criticalActions` | Plan → YAML (mandatory) |
| Workflow Paths | `skills` | Module-specific |
| Sidecar Need | `sidecar` | If multi-workflow |
**MODULE-SPECIAL ENHANCEMENTS:**
```yaml
# Module agents include workflow integration
skills:
- workflow: "{project-root}/_bmad/{module-id}/workflows/{workflow-name}/workflow.md"
description: "From plan workflow list"
- workflow: "{project-root}/_bmad/{module-id}/workflows/{another-workflow}/workflow.md"
description: "From plan workflow list"
# Optional: Sidecar for complex modules
sidecar:
enabled: true
workflows:
- ref: "primary-workflow"
type: "primary"
- ref: "secondary-workflow"
type: "support"
```
**CRITICAL ACTIONS MAPPING:**
```
For each critical action in plan:
1. Identify matching command in YAML
2. Add `critical: true` flag
3. Ensure handler references agent function
```
### 5. Create Sidecar (If Needed)
**SIDEAR REQUIRED IF:**
- Module has 3+ workflows
- Workflows have complex interdependencies
- Module needs initialization workflow
**SIDECAR STRUCTURE:**
```yaml
# {agent-name}.sidecar.yaml
sidecar:
module: "{module-id}"
initialization:
workflow: "workflow-init"
required: true
workflows:
- name: "workflow-name"
path: "workflows/{workflow-name}/workflow.md"
type: "primary|support|utility"
dependencies: []
agent:
path: "{agent-name}.agent.yaml"
```
**IF SIDEAR NOT NEEDED**: Skip this step
### 6. Write Module Agent YAML
```
Write: agentYamlOutput (using dynamic path)
Content: Assembled YAML from step 4
```
**Validation Checklist:**
- [ ] All plan fields present in YAML
- [ ] Workflow paths are valid and correct
- [ ] Critical actions flagged
- [ ] Sidecar created (if needed) or skipped (if not)
- [ ] YAML syntax is valid
- [ ] Language choice preserved throughout
## COMPLETION PHASE
### 7. Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Write agent YAML to {agentBuildOutput}/{agent-name}/{agent-name}.agent.yaml (or appropriate output path), update frontmatter, then only then load, read entire file, then execute {nextStepFile}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#7-present-menu-options)
#### EXECUTION RULES:
- ALWAYS halt and wait for user input after presenting menu
- ONLY proceed to next step when user selects 'C'
- After other menu items execution, return to this menu
- User can chat or ask questions - always respond and then end with display again of the menu options
**USER RESPONSE HANDLING:**
- **Option 1**: Proceed to step-07a-plan-traceability.md with sequential mode
- **Option 2**: Proceed to step-07a-plan-traceability.md with yolo mode
- **Invalid input**: Re-ask with options
# CONTEXT BOUNDARIES
**IN SCOPE:**
- Reading expert template and architecture
- Loading agent plan
- Assembling Module agent YAML
- Creating sidecar (if needed)
- Writing valid YAML output
**OUT OF SCOPE:**
- Modifying plan content
- Creating new template structures
- Implementing agent code
- Writing workflow files
- Testing agent functionality
**DO NOT:**
- Add commands not in plan
- Modify persona from plan
- Create custom template structures
- Skip critical actions mapping
- Assume sidecar need - evaluate based on workflow count
# CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [complete YAML generated and written to output], will you then load and read fully `{nextStepFile}` to execute and celebrate completion.
**THIS STEP IS COMPLETE WHEN:**
1. Module agent YAML file exists at agentYamlOutput path
2. YAML contains all plan content correctly mapped
3. Module-specific workflow paths are configured
4. Sidecar is created (if needed) or correctly skipped (if not)
5. User has chosen review mode (one-at-a-time or YOLO)
6. Ready to proceed to step-07a-plan-traceability.md
**STOP BEFORE:**
- Writing workflow implementations
- Creating agent code files
- Testing agent functionality
- Deploying to active system
# SUCCESS METRICS
**COMPLETION:**
- [ ] Module agent YAML exists with all required fields
- [ ] All plan content accurately mapped to YAML
- [ ] Workflow integration paths configured correctly
- [ ] Critical actions properly flagged
- [ ] Sidecar created or correctly skipped
- [ ] YAML syntax is valid
- [ ] User confirms review mode choice
- [ ] Transitions to step-07a-plan-traceability.md
**VALIDATION:**
- Plan-to-YAML mapping: 100% accuracy
- Workflow paths: All valid and correct
- Critical actions: All present and flagged
- Sidecar decision: Correctly evaluated
- Language choice: Preserved throughout
# FAILURE MODES
**IF PLAN MISSING CONTENT:**
→ Return to step-02-discover.md to complete plan
**IF EXPERT TEMPLATE MISSING:**
→ Raise error - template is mandatory baseline
**IF YAML SYNTAX ERROR:**
→ Fix and retry write operation
**IF WORKFLOW PATHS INVALID:**
→ Flag for review in traceability step
**IF USER ASKS FOR MODIFICATIONS:**
→ Return to appropriate planning step (03-persona, 04-commands, or 05-name)
_bmad/bmb/workflows/agent/steps-e/e-01-load-existing.md
+1 -1
View File
@@ -4,7 +4,7 @@ description: 'Load and analyze existing agent for editing'
# File References
thisStepFile: ./e-01-load-existing.md
workflowFile: ../workflow.md
workflowFile: ../workflow-edit-agent.md
nextStepFile: './e-02-discover-edits.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
agentMetadata: ../data/agent-metadata.md
_bmad/bmb/workflows/agent/steps-e/e-02-discover-edits.md
+9 -8
View File
@@ -2,7 +2,7 @@
name: 'e-02-discover-edits'
description: 'Discover what user wants to change about the agent'
nextStepFile: './e-04-type-metadata.md'
nextStepFile: './e-04-sidecar-metadata.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
@@ -74,8 +74,8 @@ I can help you modify:
**[P]ersona** - Role, identity, communication style, principles
**[C]ommands** - Add, remove, or modify commands and menu structure
**[M]etadata** - Name, description, version, tags, category
**[S]idecar** - Add or remove memory (convert hasSidecar: true/false)
**[A]ctions** - Critical actions and activation behaviors
**[T]ype** - Convert between Simple/Expert/Module types
**[O]ther** - Configuration, capabilities, system context
Which areas would you like to edit? (You can select multiple)"
@@ -107,10 +107,10 @@ Which areas would you like to edit? (You can select multiple)"
- "Should new activation behaviors be added?"
- "Are current actions executing as expected?"
#### If Type conversion selected:
- "What type are you converting from/to?"
- "What's driving this conversion?"
- "Are you aware of the implications (e.g., Expert needs sidecar)?"
#### If Sidecar selected:
- "Do you want to add memory (hasSidecar: true) or remove it (hasSidecar: false)?"
- "What should the agent remember across sessions?"
- "Are you aware of the implications?"
### 4. Document Edits to Plan
@@ -135,8 +135,9 @@ Which areas would you like to edit? (You can select multiple)"
- [ ] {edit description}
- [ ] {edit description}
### Type Conversion
- [ ] {from: X, to: Y, rationale: ...}
### Sidecar Conversion
- [ ] {from: hasSidecar: false, to: hasSidecar: true, rationale: ...}
- [ ] {from: hasSidecar: true, to: hasSidecar: false, rationale: ...}
### Other Edits
- [ ] {edit description}
_bmad/bmb/workflows/agent/steps-e/e-04-type-metadata.md → _bmad/bmb/workflows/agent/steps-e/e-04-sidecar-metadata.md
+18 -17
View File
@@ -1,5 +1,5 @@
---
name: 'e-04-type-metadata'
name: 'e-04-sidecar-metadata'
description: 'Review and plan metadata edits'
nextStepFile: './e-05-persona.md'
@@ -11,11 +11,11 @@ advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitati
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# Edit Step 4: Type and Metadata
# Edit Step 4: Sidecar and Metadata
## STEP GOAL:
Review the agent's type and metadata, and plan any changes. If edits involve type conversion, identify the implications.
Review the agent's hasSidecar decision and metadata, and plan any changes. If edits involve sidecar conversion, identify the implications.
## MANDATORY EXECUTION RULES:
@@ -26,7 +26,7 @@ Review the agent's type and metadata, and plan any changes. If edits involve typ
### Step-Specific Rules:
- 🎯 Load reference documents before discussing edits
- 📊 Document type conversion requirements if applicable
- 📊 Document sidecar conversion requirements if applicable
- 💬 Focus on metadata that user wants to change
## EXECUTION PROTOCOLS:
@@ -42,22 +42,24 @@ Review the agent's type and metadata, and plan any changes. If edits involve typ
### 1. Load Reference Documents
Read `{agentMetadata}` and `{agentTypesDoc}` to understand validation rules and type implications.
Read `{agentMetadata}` and `{agentTypesDoc}` to understand validation rules and sidecar implications.
### 2. Review Current Metadata
From `{editPlan}`, display current:
- agentType (simple/expert/module)
- All metadata fields: id, name, title, icon, module, hasSidecar
- hasSidecar (true/false)
- All metadata fields: id, name, title, icon, module
### 3. Discuss Metadata Edits
If user wants metadata changes:
**For type conversion:**
- "Converting from {current} to {target}"
- Explain implications (e.g., Simple → Expert requires sidecar)
- Update editPlan with type conversion
**For sidecar conversion:**
- "Converting from hasSidecar: {current} to {target}"
- Explain implications:
- false → true: Need to create sidecar folder, add critical_actions with sidecar file loading
- true → false: Remove sidecar fields; if critical_actions only has sidecar references, remove section; otherwise keep non-sidecar critical_actions
- Update editPlan with conversion
**For metadata field changes:**
- id: kebab-case requirements
@@ -65,7 +67,6 @@ If user wants metadata changes:
- title: function description format
- icon: emoji/symbol
- module: path format
- hasSidecar: boolean implications
### 4. Document to Edit Plan
@@ -73,9 +74,9 @@ Append to `{editPlan}`:
```yaml
metadataEdits:
typeConversion:
from: {current-type}
to: {target-type}
sidecarConversion:
from: {current-hasSidecar}
to: {target-hasSidecar}
rationale: {explanation}
fieldChanges:
- field: {field-name}
@@ -112,13 +113,13 @@ ONLY WHEN [C continue option] is selected and [metadata changes documented], wil
- Reference documents loaded
- Metadata changes discussed and documented
- Type conversion implications understood
- Sidecar conversion implications understood
- Edit plan updated
### ❌ SYSTEM FAILURE:
- Proceeded without loading reference documents
- Type conversion without understanding implications
- Sidecar conversion without understanding implications
- Changes not documented to edit plan
**Master Rule:** Skipping steps, optimizing sequences, or not following exact instructions is FORBIDDEN and constitutes SYSTEM FAILURE.
_bmad/bmb/workflows/agent/steps-e/e-07-activation.md
+27 -29
View File
@@ -1,14 +1,12 @@
---
name: 'e-07-activation'
description: 'Review critical_actions and route to type-specific edit'
description: 'Review critical_actions and route to edit step'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
criticalActions: ../data/critical-actions.md
# Type-specific edit routes
simpleEdit: './e-08a-edit-simple.md'
expertEdit: './e-08b-edit-expert.md'
moduleEdit: './e-08c-edit-module.md'
# Edit step route (determined by hasSidecar)
agentEdit: './e-08-edit-agent.md'
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
@@ -18,26 +16,26 @@ partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
## STEP GOAL:
Review critical_actions and route to the appropriate type-specific edit step (Simple/Expert/Module).
Review critical_actions and route to the agent edit step based on hasSidecar value.
## MANDATORY EXECUTION RULES:
- 📖 CRITICAL: Read the complete step file before taking any action
- 🔄 CRITICAL: Load criticalActions and editPlan first
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}}`
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Load criticalActions.md before discussing activation
- 📊 Determine target type for routing
- 💬 Route based on POST-EDIT agent type
- 📊 Determine hasSidecar for routing
- 💬 Route based on POST-EDIT hasSidecar value
## EXECUTION PROTOCOLS:
- 🎯 Load criticalActions.md
- 📊 Check editPlan for target agent type
- 💾 Route to appropriate type-specific edit step
- ➡️ Auto-advance to type-specific edit on [C]
- 📊 Check editPlan for target hasSidecar value
- 💾 Route to agent edit step
- ➡️ Auto-advance to edit step on [C]
## MANDATORY SEQUENCE
@@ -47,26 +45,27 @@ Review critical_actions and route to the appropriate type-specific edit step (Si
Read `{criticalActions}` and `{editPlan}` to understand:
- Current critical_actions (if any)
- Target agent type after edits
- Target hasSidecar value after edits
### 2. Review Critical Actions
If user wants to add/modify critical_actions:
- Reference patterns from criticalActions.md
- Define action name, description, invocation
- For Expert agents: specify sidecar-folder and file paths
- For hasSidecar: true — specify sidecar-folder and file paths
### 3. Determine Routing
Check `{editPlan}` for agent metadata (module and hasSidecar):
Check `{editPlan}` for agent metadata (hasSidecar):
```yaml
# Determine agent type from module + hasSidecar combination
module ≠ "stand-alone" → route to e-08c-edit-module.md
module = "stand-alone" + hasSidecar: true → route to e-08b-edit-expert.md
module = "stand-alone" + hasSidecar: false → route to e-08a-edit-simple.md
# Simple routing based on hasSidecar
hasSidecar: true → route to e-08-edit-agent.md (create sidecar structure)
hasSidecar: false → route to e-08-edit-agent.md (single YAML file)
```
The edit step handles both cases based on hasSidecar value.
### 4. Document to Edit Plan
Append to `{editPlan}`:
@@ -77,19 +76,19 @@ activationEdits:
additions: []
modifications: []
routing:
destinationEdit: {e-08a|e-08b|e-08c}
sourceType: {simple|expert|module} # Derived from module + hasSidecar
destinationEdit: e-08-edit-agent.md
hasSidecar: {true|false} # Derived from edit plan
```
### 5. Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue to Type-Specific Edit"
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue to Edit Agent"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Save to {editPlan}, determine routing based on module + hasSidecar, then only then load and execute the appropriate type-specific edit step
- IF C: Save to {editPlan}, then only then load and execute the agent edit step
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#5-present-menu-options)
#### EXECUTION RULES:
@@ -100,11 +99,10 @@ Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Cont
## CRITICAL STEP COMPLETION NOTE
This is the **ROUTING HUB** for edit flow. ONLY WHEN [C continue option] is selected and [routing determined], load and execute the appropriate type-specific edit step:
This is the **ROUTING HUB** for edit flow. ONLY WHEN [C continue option] is selected and [routing determined], load and execute the agent edit step:
- module ≠ "stand-alone" → e-08c-edit-module.md (Module agent)
- module = "stand-alone" + hasSidecar: true → e-08b-edit-expert.md (Expert agent)
- module = "stand-alone" + hasSidecar: false → e-08a-edit-simple.md (Simple agent)
- hasSidecar: false → Single YAML file edit
- hasSidecar: true → YAML + sidecar folder structure edit
---
@@ -113,13 +111,13 @@ This is the **ROUTING HUB** for edit flow. ONLY WHEN [C continue option] is sele
### ✅ SUCCESS:
- criticalActions.md loaded
- Routing determined based on target type
- Routing determined based on hasSidecar
- Edit plan updated with routing info
### ❌ SYSTEM FAILURE:
- Proceeded without loading reference documents
- Routing not determined
- Wrong type-specific edit step selected
- Wrong edit step selected
**Master Rule:** Skipping steps, optimizing sequences, or not following exact instructions is FORBIDDEN and constitutes SYSTEM FAILURE.
_bmad/bmb/workflows/agent/steps-e/e-08-edit-agent.md
+196
View File
@@ -0,0 +1,196 @@
---
name: 'e-08-edit-agent'
description: 'Apply edits to agent (with or without sidecar)'
nextStepFile: './e-09-celebrate.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
agentFile: '{original-agent-path}'
agentBackup: '{original-agent-path}.backup'
# Template and Architecture
agentTemplate: ../templates/agent-template.md
agentArch: ../data/agent-architecture.md
agentValidation: ../data/agent-validation.md
agentCompilation: ../data/agent-compilation.md
agentMetadata: ../data/agent-metadata.md
personaProperties: ../data/persona-properties.md
principlesCrafting: ../data/principles-crafting.md
agentMenuPatterns: ../data/agent-menu-patterns.md
criticalActions: ../data/critical-actions.md
# Reference examples
noSidecarExample: ../data/reference/without-sidecar/commit-poet.agent.yaml
withSidecarExample: ../data/reference/with-sidecar/journal-keeper/journal-keeper.agent.yaml
advancedElicitationTask: '{project-root}/_bmad/core/workflows/advanced-elicitation/workflow.xml'
partyModeWorkflow: '{project-root}/_bmad/core/workflows/party-mode/workflow.md'
---
# Edit Step 8: Edit Agent
## STEP GOAL:
Apply all planned edits to the agent YAML file. The edit approach (with or without sidecar) is determined by the `hasSidecar` value from the edit plan.
## MANDATORY EXECUTION RULES:
- 🛑 ALWAYS create backup before modifying agent file
- 📖 CRITICAL: Read template and architecture files first
- 🔄 CRITICAL: Load editPlan and agentFile
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Load all reference files before applying edits
- 📊 Apply edits exactly as specified in editPlan
- 💾 Validate YAML after each edit
- 🎭 Handle sidecar structure if hasSidecar: true
- ➡️ Auto-advance to celebration when complete
## EXECUTION PROTOCOLS:
- 🎯 Load template, architecture, and validation files
- 📊 Read editPlan to get all planned changes
- 💾 Create backup
- 📝 Apply edits: sidecar conversion, metadata, persona, commands, critical_actions
- 🎭 Manage sidecar folder structure (if applicable)
- ✅ Validate YAML and sidecar paths
- ➡️ Auto-advance to next step
## MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
### 1. Load Reference Documents
Read all files before editing:
- `{agentTemplate}` - YAML structure reference
- `{agentArch}` - Agent architecture (with/without sidecar)
- `{agentValidation}` - Validation checklist
- `{agentCompilation}` - Assembly guidelines
- `{agentMetadata}`, `{personaProperties}`, `{principlesCrafting}`
- `{agentMenuPatterns}`, `{criticalActions}`
### 2. Load Edit Plan and Agent
Read `{editPlan}` to get all planned edits.
Read `{agentFile}` to get current agent YAML.
Check the `hasSidecar` value from editPlan to determine edit approach.
### 3. Create Backup
ALWAYS backup before editing:
```bash
cp {agentFile} {agentBackup}
```
Confirm: "Backup created at: `{agentBackup}`"
### 4. Apply Edits in Sequence
For each planned edit:
**Sidecar Conversion:**
**false → true (Adding sidecar):**
- Set `hasSidecar: true`
- Add `metadata.sidecar-folder` if not present
- Add `critical_actions` section with sidecar file references
- Create sidecar directory: `{agent-folder}/{agent-name}-sidecar/`
- Create starter files: `memories.md`, `instructions.md`
- Update all references to use `{project-root}/_bmad/_memory/{sidecar-folder}/` format
**true → false (Removing sidecar):**
- Set `hasSidecar: false`
- Remove `metadata.sidecar-folder` and `metadata.sidecar-path`
- If critical_actions contains only sidecar references, remove the section
- If critical_actions contains non-sidecar activation behaviors, keep and clean sidecar references
- Remove sidecar references from menu actions
- Optionally archive sidecar folder
**Metadata Edits:**
- Apply each field change from metadataEdits
- Validate format conventions
**Persona Edits:**
- Replace persona section with new four-field persona
- Validate field purity (role ≠ identity ≠ communication_style)
- For hasSidecar: true, ensure communication_style includes memory reference patterns
**Command Edits:**
- Additions: append to commands array
- Modifications: update specific commands
- Removals: remove from commands array
**Critical Actions Edits (hasSidecar: true only):**
- Additions: append to critical_actions array
- Modifications: update specific actions
- Removals: remove from array
- Ensure all references use correct `{project-root}/_bmad/_memory/` paths
### 5. Validate After Each Edit
**For both types:**
- Confirm YAML syntax is valid after each modification
**For hasSidecar: true:**
- Validate sidecar path format
- Ensure all critical_actions reference correct paths
- Confirm sidecar folder structure exists
### 6. Document Applied Edits
Append to `{editPlan}`:
```yaml
editsApplied:
- {edit-description}
- {edit-description}
backup: {agentBackup}
timestamp: {YYYY-MM-DD HH:MM}
```
### 7. Present MENU OPTIONS
Display: "**Select an Option:** [A] Advanced Elicitation [P] Party Mode [C] Continue"
#### Menu Handling Logic:
- IF A: Execute {advancedElicitationTask}, and when finished redisplay the menu
- IF P: Execute {partyModeWorkflow}, and when finished redisplay the menu
- IF C: Save to {editPlan}, then only then load, read entire file, then execute {nextStepFile}
- IF Any other comments or queries: help user respond then [Redisplay Menu Options](#7-present-menu-options)
#### EXECUTION RULES:
- ALWAYS halt and wait for user input after presenting menu
- ONLY proceed to next step when user selects 'C'
- After other menu items execution, return to this menu
## CRITICAL STEP COMPLETION NOTE
ONLY WHEN [C continue option] is selected and [all edits applied and validated], will you then load and read fully `{nextStepFile}` to execute and celebrate.
---
## SUCCESS METRICS
✅ Backup created
✅ All reference files loaded
✅ All edits applied correctly
✅ YAML remains valid
✅ Sidecar structure correct (if hasSidecar: true)
✅ Sidecar paths validated (if hasSidecar: true)
✅ Edit plan tracking updated
## FAILURE MODES
❌ Backup failed
❌ YAML became invalid
❌ Sidecar paths broken (hasSidecar: true)
❌ Edits not applied as specified
---
**Auto-advancing to celebration when complete...**
_bmad/bmb/workflows/agent/steps-e/e-08a-edit-simple.md
-137
View File
@@ -1,137 +0,0 @@
---
name: 'e-08a-edit-simple'
description: 'Apply edits to Simple agent'
nextStepFile: './e-09-celebrate.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
agentFile: '{original-agent-path}'
agentBackup: '{original-agent-path}.backup'
# Template and Architecture
simpleTemplate: ../templates/simple-agent.template.md
simpleArch: ../data/simple-agent-architecture.md
agentCompilation: ../data/agent-compilation.md
agentMetadata: ../data/agent-metadata.md
personaProperties: ../data/persona-properties.md
principlesCrafting: ../data/principles-crafting.md
agentMenuPatterns: ../data/agent-menu-patterns.md
criticalActions: ../data/critical-actions.md
---
# Edit Step 8a: Edit Simple Agent
## STEP GOAL:
Apply all planned edits to the Simple agent YAML file using templates and architecture references for validation.
## MANDATORY EXECUTION RULES:
- 🛑 ALWAYS create backup before modifying agent file
- 📖 CRITICAL: Read template and architecture files first
- 🔄 CRITICAL: Load editPlan and agentFile
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Load all reference files before applying edits
- 📊 Apply edits exactly as specified in editPlan
- 💾 Validate YAML after each edit
- ➡️ Auto-advance to post-edit validation when complete
## EXECUTION PROTOCOLS:
- 🎯 Load template, architecture, and data files
- 📊 Read editPlan to get all planned changes
- 💾 Create backup
- 📝 Apply edits: type conversion, metadata, persona, commands, critical_actions
- ✅ Validate YAML syntax
- ➡️ Auto-advance to next validation step
## MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
### 1. Load Reference Documents
Read all files before editing:
- `{simpleTemplate}` - YAML structure reference
- `{simpleArch}` - Simple agent architecture
- `{agentCompilation}` - Assembly guidelines
- `{agentMetadata}`, `{personaProperties}`, `{principlesCrafting}`
- `{agentMenuPatterns}`, `{criticalActions}`
### 2. Load Edit Plan and Agent
Read `{editPlan}` to get all planned edits.
Read `{agentFile}` to get current agent YAML.
### 3. Create Backup
ALWAYS backup before editing:
`cp {agentFile} {agentBackup}`
Confirm: "Backup created at: `{agentBackup}`"
### 4. Apply Edits in Sequence
For each planned edit:
**Type Conversion (Simple ← Expert/Module):**
- Converting TO Simple: Remove `metadata.sidecar-folder`, remove all sidecar references
- Set `module: stand-alone` and `hasSidecar: false`
- Remove type-specific fields from source type
**Metadata Edits:**
- Apply each field change from metadataEdits
**Persona Edits:**
- Replace persona section with new four-field persona
- Validate field purity (role ≠ identity ≠ communication_style)
**Command Edits:**
- Additions: append to commands array
- Modifications: update specific commands
- Removals: remove from commands array
**Critical Actions Edits:**
- Additions: append to critical_actions array
- Modifications: update specific actions
- Removals: remove from array
### 5. Validate YAML After Each Edit
Confirm YAML syntax is valid after each modification.
### 6. Document Applied Edits
Append to `{editPlan}`:
```yaml
editsApplied:
- {edit-description}
- {edit-description}
backup: {agentBackup}
timestamp: {YYYY-MM-DD HH:MM}
```
### 7. Auto-Advance
When all edits applied successfully, load and execute `{nextStepFile}` immediately.
## SUCCESS METRICS
✅ Backup created
✅ All reference files loaded
✅ All edits applied correctly
✅ YAML remains valid
✅ Edit plan tracking updated
## FAILURE MODES
❌ Backup failed
❌ YAML became invalid
❌ Edits not applied as specified
---
**Auto-advancing to post-edit validation...
_bmad/bmb/workflows/agent/steps-e/e-08b-edit-expert.md
-119
View File
@@ -1,119 +0,0 @@
---
name: 'e-08b-edit-expert'
description: 'Apply edits to Expert agent'
nextStepFile: './e-09-celebrate.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
agentFile: '{original-agent-path}'
agentBackup: '{original-agent-path}.backup'
# Template and Architecture
expertTemplate: ../templates/expert-agent-template/expert-agent.template.md
expertArch: ../data/expert-agent-architecture.md
agentCompilation: ../data/agent-compilation.md
agentMetadata: ../data/agent-metadata.md
personaProperties: ../data/persona-properties.md
principlesCrafting: ../data/principles-crafting.md
agentMenuPatterns: ../data/agent-menu-patterns.md
criticalActions: ../data/critical-actions.md
expertValidation: ../data/expert-agent-validation.md
---
# Edit Step 8b: Edit Expert Agent
## STEP GOAL:
Apply all planned edits to the Expert agent YAML file and manage sidecar structure changes.
## MANDATORY EXECUTION RULES:
- 🛑 ALWAYS create backup before modifying agent file
- 📖 CRITICAL: Read template and architecture files first
- 🔄 CRITICAL: Load editPlan and agentFile
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Load all reference files before applying edits
- 📊 Manage sidecar structure for Expert agents
- 💾 Validate YAML and sidecar paths after edits
- ➡️ Auto-advance to post-edit validation when complete
## EXECUTION PROTOCOLS:
- 🎯 Load template, architecture, and data files
- 📊 Read editPlan to get all planned changes
- 💾 Create backup
- 📝 Apply edits including sidecar management
- ✅ Validate YAML and sidecar paths
- ➡️ Auto-advance to next validation step
## MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
### 1. Load Reference Documents
Read all files before editing:
- `{expertTemplate}` - Expert YAML structure
- `{expertArch}` - Expert agent architecture
- `{agentCompilation}`, `{agentMetadata}`, `{personaProperties}`, `{principlesCrafting}`
- `{agentMenuPatterns}`, `{criticalActions}`, `{expertValidation}`
### 2. Load Edit Plan and Agent
Read `{editPlan}` to get all planned edits.
Read `{agentFile}` to get current agent YAML.
### 3. Create Backup
ALWAYS backup before editing:
`cp {agentFile} {agentBackup}`
### 4. Apply Edits in Sequence
**Type Conversion TO Expert:**
- Set `module: stand-alone` and `hasSidecar: true`
- Add `metadata.sidecar-folder` if not present
- Create sidecar directory next to agent.yaml: `{agent-folder}/{agent-name}-sidecar/`
**Sidecar Management:**
- If changing sidecar-folder: update all critical_actions references
- If removing sidecar (Expert → Simple): remove sidecar fields and folder
- Create/update sidecar files as needed
**Metadata, Persona, Commands, Critical Actions:**
- Same as Simple agent edit
### 5. Validate Sidecar Paths
After editing, confirm all critical_actions reference correct sidecar paths:
`{project-root}/_bmad/_memory/{sidecar-folder}/{file}.md`
### 6. Document Applied Edits
Append to `{editPlan}` with sidecar changes noted.
### 7. Auto-Advance
When all edits applied successfully, load and execute `{nextStepFile}` immediately.
## SUCCESS METRICS
✅ Backup created
✅ All reference files loaded
✅ All edits applied correctly
✅ YAML remains valid
✅ Sidecar structure correct
✅ Sidecar paths validated
## FAILURE MODES
❌ Backup failed
❌ YAML became invalid
❌ Sidecar paths broken
❌ Edits not applied as specified
---
**Auto-advancing to post-edit validation...
_bmad/bmb/workflows/agent/steps-e/e-08c-edit-module.md
-123
View File
@@ -1,123 +0,0 @@
---
name: 'e-08c-edit-module'
description: 'Apply edits to Module agent'
nextStepFile: './e-09-celebrate.md'
editPlan: '{bmb_creations_output_folder}/edit-plan-{agent-name}.md'
agentFile: '{original-agent-path}'
agentBackup: '{original-agent-path}.backup'
# Template and Architecture (use expert as baseline for Module)
expertTemplate: ../templates/expert-agent-template/expert-agent.template.md
expertArch: ../data/expert-agent-architecture.md
moduleArch: ../data/module-agent-validation.md
agentCompilation: ../data/agent-compilation.md
agentMetadata: ../data/agent-metadata.md
personaProperties: ../data/persona-properties.md
principlesCrafting: ../data/principles-crafting.md
agentMenuPatterns: ../data/agent-menu-patterns.md
criticalActions: ../data/critical-actions.md
---
# Edit Step 8c: Edit Module Agent
## STEP GOAL:
Apply all planned edits to the Module agent YAML file and manage workflow integration and sidecar structure.
## MANDATORY EXECUTION RULES:
- 🛑 ALWAYS create backup before modifying agent file
- 📖 CRITICAL: Read template and architecture files first
- 🔄 CRITICAL: Load editPlan and agentFile
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Load all reference files before applying edits
- 📊 Manage workflow integration paths for Module agents
- 💾 Validate YAML and workflow paths after edits
- ➡️ Auto-advance to post-edit validation when complete
## EXECUTION PROTOCOLS:
- 🎯 Load template, architecture, and data files
- 📊 Read editPlan to get all planned changes
- 💾 Create backup
- 📝 Apply edits including workflow paths
- ✅ Validate YAML and workflow paths
- ➡️ Auto-advance to next validation step
## MANDATORY SEQUENCE
**CRITICAL:** Follow this sequence exactly. Do not skip, reorder, or improvise unless user explicitly requests a change.
### 1. Load Reference Documents
Read all files before editing - these are RULES that must be followed when editing agents:
- `{expertTemplate}` - Module uses expert as baseline
- `{expertArch}`, `{moduleArch}` - Architecture references
- `{agentCompilation}`, `{agentMetadata}`, `{personaProperties}`, `{principlesCrafting}`
- `{agentMenuPatterns}`, `{criticalActions}`
### 2. Load Edit Plan and Agent
Read `{editPlan}` to get all planned edits.
Read `{agentFile}` to get current agent YAML.
### 3. Create Backup
ALWAYS backup before editing:
`cp {agentFile} {agentBackup}`
### 4. Apply Edits in Sequence
**Type Conversion TO Module:**
- Set `module` to module code (e.g., `bmm`, `cis`, `bmgd`, or custom)
- Add workflow integration paths
- Optionally set `hasSidecar: true` if complex multi-workflow module
**Workflow Path Management:**
- Add: `skills: - workflow: {path}`
- Remove: delete workflow entries
- Modify: update workflow paths
**Sidecar for Multi-Workflow Modules:**
- If 3+ workflows: consider sidecar creation
- Add sidecar configuration if needed
**Metadata, Persona, Commands, Critical Actions:**
- Same as Expert agent edit
### 5. Validate Workflow Paths
After editing, confirm all workflow paths are valid:
`{project-root}/_bmad/{module-id}/workflows/{workflow-name}/workflow.md`
### 6. Document Applied Edits
Append to `{editPlan}` with workflow changes noted.
### 7. Auto-Advance
When all edits applied successfully, load and execute `{nextStepFile}` immediately.
## SUCCESS METRICS
✅ Backup created
✅ All reference files loaded
✅ All edits applied correctly
✅ YAML remains valid
✅ Workflow paths validated
✅ Sidecar structure correct (if applicable)
## FAILURE MODES
❌ Backup failed
❌ YAML became invalid
❌ Workflow paths broken
❌ Edits not applied as specified
---
**Auto-advancing to post-edit validation...
_bmad/bmb/workflows/agent/steps-v/v-01-load-review.md
+12 -11
View File
@@ -41,26 +41,27 @@ Load the existing agent file and initialize a validation report to track all fin
### 1. Load Agent File
Read the complete YAML from the agent file path provided by the user.
If the module property of the agent metadata is stand-alone, it is not a module agent.
If the module property of the agent is a module code (like bmm, bmb, etc...) it is a module agent.
If the property hasSidecar: true exists in the metadata, then it is an expert agent.
Else it is a simple agent.
If a module agent also hasSidecar: true - this means it is a modules expert agent, thus it can have sidecar.
Check the metadata to determine agent configuration:
- **module**: `stand-alone` or module code (bmm, cis, bmgd, etc.)
- **hasSidecar**: `true` or `false`
### 2. Display Agent Summary
```markdown
## Agent to Validate: {agent-name}
**Type:** {simple|expert|module}
**Configuration:** Agent {WITH|WITHOUT} sidecar
**hasSidecar:** {true|false}
**module:** {module-value}
**File:** {agent-file-path}
### Current Structure:
**Persona:** {character count} characters
**Commands:** {count} commands
**Critical Actions:** {count} actions
**Critical Actions:** {count} actions (if hasSidecar: true)
**Sidecar:** {present|not present}
```
### 3. Create Validation Report
@@ -70,7 +71,8 @@ Initialize the validation report:
```markdown
---
agentName: '{agent-name}'
agentType: '{simple|expert|module}' # Derived from module + hasSidecar
hasSidecar: {true|false}
module: '{module-value}'
agentFile: '{agent-file-path}'
validationDate: '{YYYY-MM-DD}'
stepsCompleted:
@@ -82,9 +84,8 @@ stepsCompleted:
## Agent Overview
**Name:** {agent-name}
**Type:** {simple|expert|module} # Derived from: module + hasSidecar
**module:** {module-value}
**hasSidecar:** {true|false}
**module:** {module-value}
**File:** {agent-file-path}
---
@@ -98,7 +99,7 @@ Write to `{validationReport}`.
### 4. Present MENU OPTIONS
Display: "**Is this the correct agent to validate and is it identified as the proper type?** [A] Advanced Elicitation [P] Party Mode [C] Yes, Begin Validation"
Display: "**Is this the correct agent to validate and is it identified as the proper configuration?** [A] Advanced Elicitation [P] Party Mode [C] Yes, Begin Validation"
#### Menu Handling Logic:
_bmad/bmb/workflows/agent/steps-v/v-02b-validate-persona.md
+2 -2
View File
@@ -80,7 +80,7 @@ Perform these checks systematically - validate EVERY rule specified in personaPr
- [ ] Principles are consistent with role
- [ ] 3-7 principles recommended (not overwhelming)
- [ ] Each principle is clear and specific
- [ ] First principle activates expert knowledge domain
- [ ] First principle activates domain knowledge
6. **Consistency Checks**
- [ ] Role, identity, communication_style, principles all align
@@ -101,7 +101,7 @@ Append to `{validationReport}`:
- [ ] role: specific, not generic
- [ ] identity: defines who agent is
- [ ] communication_style: speech patterns only
- [ ] principles: first principle activates expert knowledge
- [ ] principles: first principle activates domain knowledge
**Detailed Findings:**
_bmad/bmb/workflows/agent/steps-v/v-02c-validate-menu.md
+32 -50
View File
@@ -52,61 +52,41 @@ Perform these checks systematically - validate EVERY rule specified in agentMenu
- [ ] Menu section exists and is properly formatted
- [ ] At least one menu item defined (unless intentionally tool-less)
- [ ] Menu items follow proper YAML structure
- [ ] Each item has required fields (name, description, pattern)
- [ ] Each item has required fields (trigger, description, action)
2. **Menu Item Requirements**
For each menu item:
- [ ] name: Present, unique, uses kebab-case
- [ ] description: Clear and concise
- [ ] pattern: Valid regex pattern or tool reference
- [ ] scope: Appropriate scope defined (if applicable)
- [ ] trigger: Present, follows `XX or fuzzy match on command` format
- [ ] description: Clear and concise, starts with `[XX]` code
- [ ] action: Prompt reference (#id) or inline instruction
3. **Pattern Quality**
- [ ] Patterns are valid and testable
- [ ] Patterns are specific enough to match intended inputs
- [ ] Patterns are not overly restrictive
- [ ] Patterns use appropriate regex syntax
3. **Trigger Format Validation**
- [ ] Format: `XX or fuzzy match on command-name` (XX = 2-letter code)
- [ ] Codes are unique within agent
- [ ] No reserved codes used: MH, CH, PM, DA
4. **Description Quality**
- [ ] Each item has clear description
- [ ] Descriptions explain what the item does
- [ ] Descriptions are consistent in style
- [ ] Descriptions help users understand when to use
4. **Description Format Validation**
- [ ] Descriptions start with `[XX]` code
- [ ] Code in description matches trigger code
- [ ] Descriptions are clear and descriptive
5. **Alignment Checks**
5. **Action Handler Validation**
- [ ] If `action: '#prompt-id'`, corresponding prompt exists
- [ ] If `action: 'inline text'`, instruction is complete and clear
6. **Alignment Checks**
- [ ] Menu items align with agent's role/purpose
- [ ] Menu items are supported by agent's expertise
- [ ] Menu items fit within agent's constraints
- [ ] Menu items are appropriate for target users
6. **Completeness**
- [ ] Core capabilities for this role are covered
- [ ] No obvious missing functionality
- [ ] Menu scope is appropriate (not too sparse/overloaded)
- [ ] Related functionality is grouped logically
7. **Standards Compliance**
- [ ] No prohibited patterns or commands
- [ ] No security vulnerabilities in patterns
- [ ] No ambiguous or conflicting items
- [ ] Consistent naming conventions
8. **Menu Link Validation (Agent Type Specific)**
- [ ] Determine agent type from metadata:
- Simple: module property is 'stand-alone' AND hasSidecar is false/absent
- Expert: hasSidecar is true
- Module: module property is a module code (e.g., 'bmm', 'bmb', 'bmgd', 'bmad')
- [ ] For Expert agents (hasSidecar: true):
- Menu handlers SHOULD reference external sidecar files (e.g., `./{agent-name}-sidecar/...`)
- OR have inline prompts defined directly in the handler
- [ ] For Module agents (module property is a module code):
- Menu handlers SHOULD reference external module files under the module path
- Exec paths must start with `{project-root}/_bmad/{module}/...`
- Verify referenced files exist under the module directory
- [ ] For Simple agents (stand-alone, no sidecar):
- Menu handlers MUST NOT have external file links
- Menu handlers SHOULD only use relative links within the same file (e.g., `#section-name`)
- OR have inline prompts defined directly in the handler
7. **Configuration Specific Menu Handler Validation**
- [ ] Determine hasSidecar from metadata
- [ ] For hasSidecar: true:
- [ ] Menu handlers MAY reference sidecar files using correct path format
- [ ] Sidecar references use: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] For hasSidecar: false:
- [ ] Menu handlers MUST NOT have sidecar file links
- [ ] Menu handlers use only internal references (#) or inline prompts
### 3. Append Findings to Report
@@ -117,12 +97,14 @@ Append to `{validationReport}`:
**Status:** {✅ PASS / ⚠️ WARNING / ❌ FAIL}
**hasSidecar:** {true|false}
**Checks:**
- [ ] A/P/C convention followed
- [ ] Command names clear and descriptive
- [ ] Command descriptions specific and actionable
- [ ] Menu handling logic properly specified
- [ ] Agent type appropriate menu links verified
- [ ] Triggers follow `XX or fuzzy match on command` format
- [ ] Descriptions start with `[XX]` code
- [ ] No reserved codes (MH, CH, PM, DA)
- [ ] Action handlers valid (#prompt-id or inline)
- [ ] Configuration appropriate menu links
**Detailed Findings:**
_bmad/bmb/workflows/agent/steps-v/v-02d-validate-structure.md
+21 -23
View File
@@ -4,8 +4,7 @@ description: 'Validate YAML structure and append to report'
nextStepFile: './v-02e-validate-sidecar.md'
validationReport: '{bmb_creations_output_folder}/validation-report-{agent-name}.md'
simpleValidation: ../data/simple-agent-validation.md
expertValidation: ../data/expert-agent-validation.md
agentValidation: ../data/agent-validation.md
agentCompilation: ../data/agent-compilation.md
agentFile: '{agent-file-path}'
---
@@ -44,7 +43,7 @@ Validate the agent's YAML structure and completeness against BMAD standards as d
### 1. Load References
Read `{agentCompilation}`, `{simpleValidation}`, `{expertValidation}`, `{validationReport}`, and `{agentFile}`.
Read `{agentCompilation}`, `{agentValidation}`, `{validationReport}`, and `{agentFile}`.
### 2. Validate Structure
@@ -64,7 +63,7 @@ Perform these checks systematically - validate EVERY rule specified in agentComp
- [ ] Boolean fields are actual booleans (not strings)
#### C. Section Completeness
- [ ] All required sections present based on agent type
- [ ] All required sections present based on hasSidecar value
- [ ] Sections not empty unless explicitly optional
- [ ] Proper markdown heading hierarchy (##, ###)
- [ ] No orphaned content without section headers
@@ -75,27 +74,24 @@ Perform these checks systematically - validate EVERY rule specified in agentComp
- [ ] No malformed YAML structures
- [ ] File references use correct path format
#### E. Agent Type Specific Checks
#### E. Agent Configuration Specific Checks
**For Simple Agents (hasSidecar is false/absent, module is 'stand-alone'):**
**For Agents WITHOUT Sidecar (hasSidecar is false):**
- [ ] No sidecar requirements
- [ ] No sidecar-folder path in metadata
- [ ] Basic fields complete
- [ ] No expert-only configuration present
- [ ] If critical_actions present, no sidecar file references
- [ ] Menu handlers use only internal references (#) or inline prompts
- [ ] Total size under ~250 lines (unless justified)
**For Expert Agents (hasSidecar is true):**
- [ ] Sidecar flag set correctly in metadata
**For Agents WITH Sidecar (hasSidecar is true):**
- [ ] hasSidecar flag set correctly in metadata
- [ ] Sidecar folder path specified in metadata
- [ ] All expert fields present
- [ ] Advanced features properly configured
- [ ] Menu handlers reference sidecar files or have inline prompts
**For Module Agents (module is a module code like 'bmm', 'bmb', etc.):**
- [ ] Module property is valid module code
- [ ] Exec paths for menu handlers start with `{project-root}/_bmad/{module}/...`
- [ ] Referenced files exist under the module directory
- [ ] If also hasSidecar: true, sidecar configuration is valid
- [ ] critical_actions section present with minimum requirements:
- [ ] Loads sidecar memories
- [ ] Loads sidecar instructions
- [ ] Restricts file access to sidecar folder
- [ ] All critical_actions reference correct `{project-root}/_bmad/_memory/` paths
- [ ] Menu handlers that update sidecar use correct path format
### 3. Append Findings to Report
@@ -106,14 +102,16 @@ Append to `{validationReport}`:
**Status:** {✅ PASS / ⚠️ WARNING / ❌ FAIL}
**Agent Type:** {simple|expert|module}
**Configuration:** Agent {WITH|WITHOUT} sidecar
**hasSidecar:** {true|false}
**Checks:**
- [ ] Valid YAML syntax
- [ ] Required fields present (name, description, type, persona)
- [ ] Field types correct (arrays, strings)
- [ ] Required fields present (name, description, persona, menu)
- [ ] Field types correct (arrays, strings, booleans)
- [ ] Consistent 2-space indentation
- [ ] Agent type appropriate structure
- [ ] Configuration appropriate structure
**Detailed Findings:**
_bmad/bmb/workflows/agent/steps-v/v-02e-validate-sidecar.md
+32 -34
View File
@@ -4,7 +4,7 @@ description: 'Validate sidecar structure and append to report'
nextStepFile: './v-03-summary.md'
validationReport: '{bmb_creations_output_folder}/validation-report-{agent-name}.md'
expertValidation: ../data/expert-agent-validation.md
agentValidation: ../data/agent-validation.md
criticalActions: ../data/critical-actions.md
agentFile: '{agent-file-path}'
sidecarFolder: '{agent-sidecar-folder}'
@@ -14,27 +14,27 @@ sidecarFolder: '{agent-sidecar-folder}'
## STEP GOAL
Validate the agent's sidecar structure (if Expert type) against BMAD standards as defined in expertValidation.md. Append findings to validation report and auto-advance.
Validate the agent's sidecar structure (if hasSidecar: true) against BMAD standards as defined in agentValidation.md. Append findings to validation report and auto-advance.
## MANDATORY EXECUTION RULES
- 📖 CRITICAL: Read the complete step file before taking any action
- 🔄 CRITICAL: Read validationReport and expertValidation first
- 🔄 CRITICAL: Read validationReport and agentValidation first
- 🔄 CRITICAL: Load the actual agent file to check for sidecar
- 🚫 NO MENU - append findings and auto-advance
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### Step-Specific Rules:
- 🎯 Validate sidecar against expertValidation.md rules (for Expert agents)
- 🎯 Validate sidecar against agentValidation.md rules (for agents with sidecar)
- 📊 Append findings to validation report
- 🚫 FORBIDDEN to present menu
## EXECUTION PROTOCOLS
- 🎯 Load expertValidation.md reference
- 🎯 Load agentValidation.md reference
- 🎯 Load the actual agent file for validation
- 📊 Validate sidecar if Expert type, skip for Simple/Module
- 📊 Validate sidecar if hasSidecar: true, skip for hasSidecar: false
- 💾 Append findings to validation report
- ➡️ Auto-advance to summary step
@@ -44,12 +44,12 @@ Validate the agent's sidecar structure (if Expert type) against BMAD standards a
### 1. Load References
Read `{expertValidation}`, `{criticalActions}`, `{validationReport}`, and `{agentFile}`.
Read `{agentValidation}`, `{criticalActions}`, `{validationReport}`, and `{agentFile}`.
### 2. Conditional Validation
**IF (module = "stand-alone" AND hasSidecar = true) OR (module ≠ "stand-alone" AND hasSidecar = true):**
Perform these checks systematically - validate EVERY rule specified in expertValidation.md:
**IF hasSidecar = true:**
Perform these checks systematically - validate EVERY rule specified in agentValidation.md:
#### A. Sidecar Folder Validation
- [ ] Sidecar folder exists at specified path
@@ -59,39 +59,37 @@ Perform these checks systematically - validate EVERY rule specified in expertVal
#### B. Sidecar File Inventory
- [ ] List all files in sidecar folder
- [ ] Verify expected files are present
- [ ] Verify expected files are present (memories.md, instructions.md recommended)
- [ ] Check for unexpected files
- [ ] Validate file names follow conventions
#### C. Path Reference Validation
For each sidecar path reference in agent YAML:
- [ ] Extract path from YAML reference
- [ ] Verify file exists at referenced path
- [ ] Check path format is correct (relative/absolute as expected)
- [ ] Verify path format is correct: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] `{project-root}` is literal
- [ ] `{sidecar-folder}` is actual folder name
- [ ] Validate no broken path references
#### D. Critical Actions File Validation (if present)
- [ ] critical-actions.md file exists
- [ ] File has proper frontmatter
- [ ] Actions section is present and not empty
- [ ] No critical sections missing
- [ ] File content is complete (not just placeholder)
#### D. Critical Actions Validation (MANDATORY for hasSidecar: true)
- [ ] critical_actions section exists in agent YAML
- [ ] Contains at minimum 3 actions
- [ ] Loads sidecar memories: `{project-root}/_bmad/_memory/{sidecar-folder}/memories.md`
- [ ] Loads sidecar instructions: `{project-root}/_bmad/_memory/{sidecar-folder}/instructions.md`
- [ ] Restricts file access: `ONLY read/write files in {project-root}/_bmad/_memory/{sidecar-folder}/`
- [ ] No placeholder text in critical_actions
- [ ] No compiler-injected steps
#### E. Module Files Validation (if present)
- [ ] Module files exist at referenced paths
- [ ] Each module file has proper frontmatter
- [ ] Module file content is complete
- [ ] No empty or placeholder module files
#### F. Sidecar Structure Completeness
#### E. Sidecar Structure Completeness
- [ ] All referenced sidecar files present
- [ ] No orphaned references (files referenced but not present)
- [ ] No unreferenced files (files present but not referenced)
- [ ] File structure matches expert agent requirements
- [ ] File structure matches agent requirements
**IF (module = "stand-alone" AND hasSidecar = false):**
**IF hasSidecar = false:**
- [ ] Mark sidecar validation as N/A
- [ ] Confirm no sidecar-folder path in metadata
- [ ] Confirm no sidecar references in critical_actions (if present)
- [ ] Confirm no sidecar references in menu handlers
### 3. Append Findings to Report
@@ -103,18 +101,18 @@ Append to `{validationReport}`:
**Status:** {✅ PASS / ⚠️ WARNING / ❌ FAIL / N/A}
**Agent Type:** {simple|expert|module with sidecar}
**hasSidecar:** {true|false}
**Checks:**
- [ ] metadata.sidecar-folder present (Expert only)
- [ ] sidecar-path format correct
- [ ] Sidecar files exist at specified path
- [ ] metadata.sidecar-folder present (if hasSidecar: true)
- [ ] Sidecar path format correct: `{project-root}/_bmad/_memory/{sidecar-folder}/...`
- [ ] Sidecar files exist at specified path (if hasSidecar: true)
- [ ] All referenced files present
- [ ] No broken path references
**Detailed Findings:**
*PASSING (for Expert agents):*
*PASSING (for agents WITH sidecar):*
{List of passing checks}
*WARNINGS:*
@@ -123,8 +121,8 @@ Append to `{validationReport}`:
*FAILURES:*
{List of blocking issues that must be fixed}
*N/A (for Simple agents):*
N/A - Agent is Simple type (module = "stand-alone" + hasSidecar: false, no sidecar required)
*N/A (for agents WITHOUT sidecar):*
N/A - Agent has hasSidecar: false, no sidecar required
```
### 4. Auto-Advance
_bmad/bmb/workflows/agent/templates/expert-agent-template/expert-agent.template.md → _bmad/bmb/workflows/agent/templates/agent-template.md
+17 -5
View File
@@ -1,8 +1,8 @@
{{#if comment}}
------------------------------------------------------------------------------
Expert Agent Handlebars Template
Used by: step-06-build.md to generate final agent YAML
Documentation: ../../data/expert-agent-architecture.md
Agent Handlebars Template (Unified)
Used by: step-07-build-agent.md to generate final agent YAML
Documentation: ../data/agent-architecture.md
------------------------------------------------------------------------------
{{/if}}
agent:
@@ -13,11 +13,15 @@ agent:
icon: {{agent_icon}}
module: {{agent_module}}{{#if agent_module_comment}} {{!-- stand-alone, bmm, cis, bmgd, or other module --}}{{/if}}
hasSidecar: {{has_sidecar}}{{#if has_sidecar_comment}} {{!-- true if agent has a sidecar folder, false otherwise --}}{{/if}}
{{#if has_sidecar}}
sidecar-folder: {{sidecar_folder}}
sidecar-path: '{project-root}/_bmad/_memory/{{sidecar_folder}}/'
{{/if}}
persona:
role: |
{{persona_role}}{{#if persona_role_note}}
{{!-- 1-2 sentences, first person --}}{{/if}}
{{!-- 1-2 sentences, first person, what the agent does --}}{{/if}}
identity: |
{{persona_identity}}{{#if persona_identity_note}}
@@ -25,17 +29,23 @@ agent:
communication_style: |
{{communication_style}}{{#if communication_style_note}}
{{!-- How the agent speaks, include memory reference patterns --}}{{/if}}
{{!-- How the agent speaks: tone, voice, mannerisms --}}
{{#if has_sidecar}}
{{!-- Include memory reference patterns: "Last time you mentioned..." or "I've noticed patterns..." --}}
{{/if}}
{{/if}}
principles:
{{#each principles}}
- {{this}}
{{/each}}
{{#if has_critical_actions}}
critical_actions:
{{#each critical_actions}}
- '{{{this}}}'
{{/each}}
{{/if}}
{{#if has_prompts}}
prompts:
@@ -51,6 +61,8 @@ agent:
- trigger: {{trigger_code}} or fuzzy match on {{trigger_command}}
{{#if action_is_prompt}}
action: '#{{action_id}}'
{{else if action_updates_sidecar}}
action: {{{action_inline}}}
{{else}}
action: {{{action_inline}}}
{{/if}}
_bmad/bmb/workflows/agent/templates/simple-agent.template.md
-72
View File
@@ -1,72 +0,0 @@
{{#if comment}}
------------------------------------------------------------------------------
Simple Agent Handlebars Template
Used by: step-06-build.md to generate final agent YAML
Documentation: ../data/simple-agent-architecture.md
------------------------------------------------------------------------------
{{/if}}
agent:
metadata:
id: {{agent_id}}
name: {{agent_name}}
title: {{agent_title}}
icon: {{agent_icon}}
module: {{agent_module}}{{#if agent_module_comment}} {{!-- stand-alone, bmm, cis, bmgd, or other module --}}{{/if}}
hasSidecar: {{has_sidecar}}{{#if has_sidecar_comment}} {{!-- true if agent has a sidecar folder, false otherwise --}}{{/if}}
persona:
role: |
{{persona_role}}{{#if persona_role_note}}
{{!-- 1-2 sentences, first person --}}{{/if}}
identity: |
{{persona_identity}}{{#if persona_identity_note}}
{{!-- 2-5 sentences, first person, background/specializations --}}{{/if}}
communication_style: |
{{communication_style}}{{#if communication_style_note}}
{{!-- How the agent speaks: tone, voice, mannerisms --}}{{/if}}
principles:
{{#each principles}}
- {{this}}
{{/each}}
{{#if has_prompts}}
prompts:
{{#each prompts}}
- id: {{id}}
content: |
{{{content}}}
{{/each}}
{{/if}}
menu:
{{#each menu_items}}
- trigger: {{trigger_code}} or fuzzy match on {{trigger_command}}
{{#if action_is_prompt}}
action: '#{{action_id}}'
{{else}}
action: {{{action_inline}}}
{{/if}}
description: '[{{trigger_code}}] {{{description}}}'
{{/each}}
{{#if has_install_config}}
install_config:
compile_time_only: true
description: '{{install_description}}'
questions:
{{#each install_questions}}
- var: {{var_name}}
prompt: '{{prompt}}'
type: {{question_type}}{{#if question_options}}
options:
{{#each question_options}}
- label: '{{label}}'
value: '{{value}}'
{{/each}}
{{/if}}
default: {{{default_value}}}
{{/each}}
{{/if}}
_bmad/bmb/workflows/agent/workflow-create-agent.md
+72
View File
@@ -0,0 +1,72 @@
---
name: create-agent
description: Create a new BMAD agent with best practices and compliance
web_bundle: true
createWorkflow: './steps-c/step-01-brainstorm.md'
---
# Create Agent
**Goal:** Collaboratively create BMAD Core compliant agents through guided discovery and systematic execution.
**Your Role:** In addition to your name, communication_style, and persona, you are also an expert agent architect specializing in BMAD Core agent creation. You guide users through creating new agents with best practices and full compliance.
---
## WORKFLOW ARCHITECTURE
This uses **step-file architecture** for disciplined execution:
### Core Principles
- **Micro-file Design**: Each step is a self-contained instruction file
- **Just-In-Time Loading**: Only the current step file is in memory
- **Sequential Enforcement**: Steps completed in order
- **State Tracking**: Document progress in tracking files (agentPlan)
- **Mode-Aware Routing**: Create-specific step flow
### Step Processing Rules
1. **READ COMPLETELY**: Always read the entire step file before taking any action
2. **FOLLOW SEQUENCE**: Execute numbered sections in order
3. **WAIT FOR INPUT**: Halt at menus and wait for user selection
4. **CHECK CONTINUATION**: Only proceed when user selects appropriate option
5. **SAVE STATE**: Update progress before loading next step
6. **LOAD NEXT**: When directed, load and execute the next step file
### Critical Rules
- 🛑 **NEVER** load multiple step files simultaneously
- 📖 **ALWAYS** read entire step file before execution
- 🚫 **NEVER** skip steps unless explicitly optional
- 💾 **ALWAYS** save progress and outputs
- 🎯 **ALWAYS** follow exact instructions in step files
- ⏸️ **ALWAYS** halt at menus and wait for input
- 📋 **NEVER** pre-load future steps
---
## INITIALIZATION SEQUENCE
### 1. Configuration Loading
Load and read full config from `{project-root}/_bmad/bmb/config.yaml`:
- `project_name`, `user_name`, `communication_language`, `document_output_language`, `bmb_creations_output_folder`
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### 2. Route to Create Workflow
"**Create Mode: Building a new BMAD Core compliant agent from scratch.**"
Load, read completely, then execute `{createWorkflow}` (steps-c/step-01-brainstorm.md)
---
## CREATE MODE NOTES
- Starts with optional brainstorming
- Progresses through discovery, metadata, persona, commands, activation
- Builds agent based on type (Simple/Expert/Module)
- Validates built agent
- Celebrates completion with installation guidance
_bmad/bmb/workflows/agent/workflow-edit-agent.md
+75
View File
@@ -0,0 +1,75 @@
---
name: edit-agent
description: Edit existing BMAD agents while maintaining compliance
web_bundle: true
editWorkflow: './steps-e/e-01-load-existing.md'
---
# Edit Agent
**Goal:** Modify existing BMAD Core compliant agents while maintaining their integrity and compliance.
**Your Role:** In addition to your name, communication_style, and persona, you are also an expert agent architect specializing in BMAD Core agent lifecycle management. You guide users through editing existing agents while preserving their core functionality and compliance.
---
## WORKFLOW ARCHITECTURE
This uses **step-file architecture** for disciplined execution:
### Core Principles
- **Micro-file Design**: Each step is a self-contained instruction file
- **Just-In-Time Loading**: Only the current step file is in memory
- **Sequential Enforcement**: Steps completed in order
- **State Tracking**: Document progress in tracking files (editPlan)
- **Mode-Aware Routing**: Edit-specific step flow
### Step Processing Rules
1. **READ COMPLETELY**: Always read the entire step file before taking any action
2. **FOLLOW SEQUENCE**: Execute numbered sections in order
3. **WAIT FOR INPUT**: Halt at menus and wait for user selection
4. **CHECK CONTINUATION**: Only proceed when user selects appropriate option
5. **SAVE STATE**: Update progress before loading next step
6. **LOAD NEXT**: When directed, load and execute the next step file
### Critical Rules
- 🛑 **NEVER** load multiple step files simultaneously
- 📖 **ALWAYS** read entire step file before execution
- 🚫 **NEVER** skip steps unless explicitly optional
- 💾 **ALWAYS** save progress and outputs
- 🎯 **ALWAYS** follow exact instructions in step files
- ⏸️ **ALWAYS** halt at menus and wait for input
- 📋 **NEVER** pre-load future steps
---
## INITIALIZATION SEQUENCE
### 1. Configuration Loading
Load and read full config from `{project-root}/_bmad/bmb/config.yaml`:
- `project_name`, `user_name`, `communication_language`, `document_output_language`, `bmb_creations_output_folder`
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### 2. Route to Edit Workflow
"**Edit Mode: Modifying an existing BMAD Core compliant agent.**"
Prompt for agent file path: "Which agent would you like to edit? Please provide the path to the `.agent.yaml` file."
Then load, read completely, and execute `{editWorkflow}` (steps-e/e-01-load-existing.md)
---
## EDIT MODE NOTES
- Loads existing agent first
- Discovers what user wants to change
- Validates current agent before editing
- Creates structured edit plan
- Applies changes with validation
- Celebrates successful edit
_bmad/bmb/workflows/agent/workflow-validate-agent.md
+73
View File
@@ -0,0 +1,73 @@
---
name: validate-agent
description: Validate existing BMAD agents and offer to improve deficiencies
web_bundle: true
validateWorkflow: './steps-v/v-01-load-review.md'
---
# Validate Agent
**Goal:** Review existing BMAD Core compliant agents through systematic validation and generate comprehensive reports.
**Your Role:** In addition to your name, communication_style, and persona, you are also a validation specialist and quality assurance expert for BMAD Core agents. You conduct systematic reviews and provide actionable improvement recommendations.
---
## WORKFLOW ARCHITECTURE
This uses **step-file architecture** for disciplined execution:
### Core Principles
- **Micro-file Design**: Each step is a self-contained instruction file
- **Just-In-Time Loading**: Only the current step file is in memory
- **Sequential Enforcement**: Steps completed in order
- **State Tracking**: Document progress in tracking files (validationReport)
- **Mode-Aware Routing**: Validate-specific step flow
### Step Processing Rules
1. **READ COMPLETELY**: Always read the entire step file before taking any action
2. **FOLLOW SEQUENCE**: Execute numbered sections in order
3. **WAIT FOR INPUT**: Halt at menus and wait for user selection
4. **CHECK CONTINUATION**: Only proceed when user selects appropriate option
5. **SAVE STATE**: Update progress before loading next step
6. **LOAD NEXT**: When directed, load and execute the next step file
### Critical Rules
- 🛑 **NEVER** load multiple step files simultaneously
- 📖 **ALWAYS** read entire step file before execution
- 🚫 **NEVER** skip steps unless explicitly optional
- 💾 **ALWAYS** save progress and outputs
- 🎯 **ALWAYS** follow exact instructions in step files
- ⏸️ **ALWAYS** halt at menus and wait for input
- 📋 **NEVER** pre-load future steps
---
## INITIALIZATION SEQUENCE
### 1. Configuration Loading
Load and read full config from `{project-root}/_bmad/bmb/config.yaml`:
- `project_name`, `user_name`, `communication_language`, `document_output_language`, `bmb_creations_output_folder`
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### 2. Route to Validate Workflow
"**Validate Mode: Reviewing an existing BMAD Core compliant agent.**"
Prompt for agent file path: "Which agent would you like to validate? Please provide the path to the `.agent.yaml` file."
Then load, read completely, and execute `{validateWorkflow}` (steps-v/v-01-load-review.md)
---
## VALIDATE MODE NOTES
- Loads existing agent
- Runs systematic validation (metadata, persona, menu, structure, sidecar)
- Generates comprehensive validation report
- Offers option to apply fixes if user desires
_bmad/bmb/workflows/agent/workflow.md
-123
View File
@@ -1,123 +0,0 @@
---
name: agent
description: Tri-modal workflow for creating, editing, and validating BMAD Core compliant agents
web_bundle: true
---
# Agent Workflow
**Goal:** Collaboratively create, edit, or validate BMAD Core compliant agents through guided discovery and systematic execution.
**Your Role:** In addition to your name, communication_style, and persona, you are also an expert agent architect specializing in BMAD Core agent lifecycle management. You guide users through creating new agents, editing existing ones, or validating agent configurations.
---
## WORKFLOW ARCHITECTURE
This uses **step-file architecture** for disciplined execution:
### Core Principles
- **Micro-file Design**: Each step is a self-contained instruction file
- **Just-In-Time Loading**: Only the current step file is in memory
- **Sequential Enforcement**: Steps completed in order, conditional based on mode
- **State Tracking**: Document progress in tracking files (agentPlan, editPlan, validationReport)
- **Mode-Aware Routing**: Separate step flows for Create/Edit/Validate
### Step Processing Rules
1. **READ COMPLETELY**: Always read the entire step file before taking any action
2. **FOLLOW SEQUENCE**: Execute numbered sections in order
3. **WAIT FOR INPUT**: Halt at menus and wait for user selection
4. **CHECK CONTINUATION**: Only proceed when user selects appropriate option
5. **SAVE STATE**: Update progress before loading next step
6. **LOAD NEXT**: When directed, load and execute the next step file
### Critical Rules
- 🛑 **NEVER** load multiple step files simultaneously
- 📖 **ALWAYS** read entire step file before execution
- 🚫 **NEVER** skip steps unless explicitly optional
- 💾 **ALWAYS** save progress and outputs
- 🎯 **ALWAYS** follow exact instructions in step files
- ⏸️ **ALWAYS** halt at menus and wait for input
- 📋 **NEVER** pre-load future steps
---
## MODE OVERVIEW
This workflow supports three modes:
| Mode | Purpose | Entry Point | Output |
|------|---------|-------------|--------|
| **Create** | Build new agent from scratch | `steps-c/step-01-brainstorm.md` | New `.agent.yaml` file |
| **Edit** | Modify existing agent | `steps-e/e-01-load-existing.md` | Updated `.agent.yaml` file |
| **Validate** | Review existing agent | `steps-v/v-01-load-review.md` | Validation report |
---
## INITIALIZATION SEQUENCE
### 1. Configuration Loading
Load and read full config from `{project-root}/_bmad/bmb/config.yaml`:
- `project_name`, `user_name`, `communication_language`, `document_output_language`, `bmb_creations_output_folder`
- ✅ YOU MUST ALWAYS SPEAK OUTPUT In your Agent communication style with the config `{communication_language}`
### 2. Mode Determination
**Check if mode was specified in the command invocation:**
- If user invoked with "create agent" or "new agent" → Set mode to **create**
- If user invoked with "edit agent" or "modify agent" → Set mode to **edit**
- If user invoked with "validate agent" or "review agent" → Set mode to **validate**
**If mode is unclear from command, ask user:**
"Welcome to the BMAD Agent Workflow! What would you like to do?
**[C]reate** - Build a new agent from scratch
**[E]dit** - Modify an existing agent
**[V]alidate** - Review an existing agent and generate report
Please select: [C]reate / [E]dit / [V]alidate"
### 3. Route to First Step
**IF mode == create:**
Load, read completely, then execute `steps-c/step-01-brainstorm.md`
**IF mode == edit:**
Prompt for agent file path: "Which agent would you like to edit? Please provide the path to the `.agent.yaml` file."
Then load, read completely, and execute `steps-e/e-01-load-existing.md`
**IF mode == validate:**
Prompt for agent file path: "Which agent would you like to validate? Please provide the path to the `.agent.yaml` file."
Then load, read completely, and execute `steps-v/v-01-load-review.md`
---
## MODE-SPECIFIC NOTES
### Create Mode
- Starts with optional brainstorming
- Progresses through discovery, metadata, persona, commands, activation
- Builds agent based on type (Simple/Expert/Module)
- Validates built agent
- Celebrates completion with installation guidance
### Edit Mode
- Loads existing agent first
- Discovers what user wants to change
- Validates current agent before editing
- Creates structured edit plan
- Applies changes with validation
- Celebrates successful edit
### Validate Mode
- Loads existing agent
- Runs systematic validation (metadata, persona, menu, structure, sidecar)
- Generates comprehensive validation report
- Offers option to apply fixes if user desires