- Removed all references to Phase 0 (should be Documentation prerequisite)
- Updated phase transitions from 'Phase 3→4' to 'Phase 3 to Phase 4'
- Ensured all phases are numbered 1-4 consistently
- Documentation for brownfield projects is now correctly referred to as 'Documentation prerequisite' rather than Phase 0
- Updated workflows-implementation.md: removed validate workflows, epic-tech-context, story-context
- Updated workflows-analysis.md: removed brainstorm-game, game-brief, added domain-research
- Updated workflows-planning.md: removed gdd, narrative, moved create-epics-and-stories to Phase 3
- Updated workflows-solutioning.md: already correct with create-epics-and-stories in Phase 3
- Removed all Mermaid diagrams and replaced with text descriptions
- Updated quick reference tables to reflect actual workflows
- Fixed flow examples to match current implementation
Copilot was triggering warning or errors in the chatmode files due to some changes in tool names.
- findTestFiles is internal tool, cannot be used.
- Other tools have change names.
- Added new tools: todos and runSubAgents.
Co-authored-by: Brian <bmadcode@gmail.com>
Add explicit radix=10 to parseInt() calls and NaN validation to prevent
unexpected hex parsing and invalid config values.
Changes:
- Line 52: Add radix and NaN check in input validation
- Line 189-192: Add radix and NaN fallback for config parsing
Fixes potential issues:
- Hex input (0x10) now rejected instead of parsed as 16
- Invalid strings return default value instead of NaN→null
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Brian <bmadcode@gmail.com>
* feat: Add provider-agnostic TTS integration via injection point system
Implements comprehensive Text-to-Speech integration for BMAD agents using a generic
TTS_INJECTION marker system. When AgentVibes (or any compatible TTS provider) is
installed, all BMAD agents can speak their responses with unique AI voices.
## Key Features
**Provider-Agnostic Architecture**
- Uses generic `TTS_INJECTION` markers instead of vendor-specific naming
- Future-proof for multiple TTS providers beyond AgentVibes
- Clean separation - BMAD stays TTS-agnostic, providers handle injection
**Installation Flow**
- BMAD → AgentVibes: TTS instructions injected when AgentVibes detects existing BMAD installation
- AgentVibes → BMAD: TTS instructions injected during BMAD installation when AgentVibes detected
- User must manually create voice assignment file when AgentVibes installs first (documented limitation)
**Party Mode Voice Support**
- Each agent speaks with unique assigned voice in multi-agent discussions
- PM, Architect, Developer, Analyst, UX Designer, etc. - all with distinct voices
**Zero Breaking Changes**
- Fully backward compatible - works without any TTS provider
- `TTS_INJECTION` markers are benign HTML comments if not processed
- No changes to existing agent behavior or non-TTS workflows
## Implementation Details
**Files Modified:**
- `tools/cli/installers/lib/core/installer.js` - TTS injection processing logic
- `tools/cli/lib/ui.js` - AgentVibes detection and installation prompts
- `tools/cli/commands/install.js` - Post-install guidance for AgentVibes setup
- `src/utility/models/fragments/activation-rules.xml` - TTS_INJECTION marker for agents
- `src/core/workflows/party-mode/instructions.md` - TTS_INJECTION marker for party mode
**Injection Point System:**
```xml
<rules>
- ALWAYS communicate in {communication_language}
<!-- TTS_INJECTION:agent-tts -->
- Stay in character until exit selected
</rules>
```
When AgentVibes is detected, the installer replaces this marker with:
```
- When responding to user messages, speak your responses using TTS:
Call: `.claude/hooks/bmad-speak.sh '{agent-id}' '{response-text}'` after each response
IMPORTANT: Use single quotes - do NOT escape special characters like ! or $
```
**Special Character Handling:**
- Explicit guidance to use single quotes without escaping
- Prevents "backslash exclamation" artifacts in speech
**User Experience:**
```
User: "How should we architect this feature?"
Architect: [Text response] + 🔊 [Professional voice explains architecture]
```
Party Mode:
```
PM (John): "I'll focus on user value..." 🔊 [Male professional voice]
UX Designer (Sara): "From a user perspective..." 🔊 [Female voice]
Architect (Marcus): "The technical approach..." 🔊 [Male technical voice]
```
## Testing
**Unit Tests:** ✅ 62/62 passing
- 49/49 schema validation tests
- 13/13 installation component tests
**Integration Testing:**
- ✅ BMAD → AgentVibes (automatic injection)
- ✅ AgentVibes → BMAD (automatic injection)
- ✅ No TTS provider (markers remain as comments)
## Documentation
Comprehensive testing guide created with:
- Both installation scenario walkthroughs
- Verification commands and expected outputs
- Troubleshooting guidance
## Known Limitations
**AgentVibes → BMAD Installation Order:**
When AgentVibes installs first, voice assignment file must be created manually:
```bash
mkdir -p .bmad/_cfg
cat > .bmad/_cfg/agent-voice-map.csv << 'EOF'
agent_id,voice_name
pm,en_US-ryan-high
architect,en_US-danny-low
dev,en_US-joe-medium
EOF
```
This limitation exists to prevent false legacy v4 detection warnings from BMAD installer.
**Recommended:** Install BMAD first, then AgentVibes for automatic voice assignment.
## Related Work
**Companion Implementation:**
- Repository: paulpreibisch/AgentVibes
- Commits: 6 commits implementing injection processing and voice routing
- Features: Retroactive injection, file path extraction, escape stripping
**GitHub Issues:**
- paulpreibisch/AgentVibes#36 - BMAD agent ID support
## Breaking Changes
None. Feature is opt-in and requires separate TTS provider installation.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Enforce project hooks over global hooks in party mode
before, claude would sometimes favor global agent vibes hooks over project specific
* feat: Automate AgentVibes installer invocation after BMAD install
Instead of showing manual installation instructions, the installer now:
- Prompts "Press Enter to start AgentVibes installer..."
- Automatically runs npx agentvibes@latest install
- Handles errors gracefully with fallback instructions
This provides a seamless installation flow matching the test script's
interactive approach.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* docs: Add automated testing script and guide for PR #934
Added comprehensive testing tools for AgentVibes party mode integration:
- test-bmad-pr.sh: Fully automated installation and verification script
- Interactive mode selection (official PR or custom fork)
- Automatic BMAD CLI setup and linking
- AgentVibes installation with guided prompts
- Built-in verification checks for voice maps and hooks
- Saved configuration for quick re-testing
- TESTING.md: Complete testing documentation
- Quick start with one-line npx command
- Manual installation alternative
- Troubleshooting guide
- Cleanup instructions
Testers can now run a single command to test the full AgentVibes integration
without needing to understand the complex setup process.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Add shell: true to npx execSync to prevent permission denied error
The execSync call for 'npx agentvibes@latest install' was failing with
'Permission denied' because the shell was trying to execute 'agentvibes@latest'
directly instead of passing it as an argument to npx.
Adding shell: true ensures the command runs in a proper shell context
where npx can correctly interpret the @latest version syntax.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Remove duplicate AgentVibes installation step from test script
The test script was calling AgentVibes installer twice:
1. BMAD installer now automatically runs AgentVibes (new feature)
2. Test script had a separate Step 6 that also ran AgentVibes
This caused the installer to run twice, with the second call failing
because it was already installed.
Changes:
- Removed redundant Step 6 (AgentVibes installation)
- Updated Step 5 to indicate it includes AgentVibes
- Updated step numbers from 7 to 6 throughout
- Added guidance that AgentVibes runs automatically
Now the flow is cleaner: BMAD installer handles everything!
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Address bmadcode review - preserve variables and move TTS logic to injection
Fixes requested changes from PR review:
1. Preserve {bmad_folder} variable placeholder
- Changed: {project_root}/.bmad/core/tasks/workflow.xml
- To: {project_root}/{bmad_folder}/core/tasks/workflow.xml
- Allows users to choose custom BMAD folder names during installation
2. Move TTS-specific hook guidance to injection system
- Removed hardcoded hook enforcement from source files
- Added hook guidance to processTTSInjectionPoints() in installer.js
- Now only appears when AgentVibes is installed (via TTS_INJECTION)
3. Maintain TTS-agnostic source architecture
- Source files remain clean of TTS-specific instructions
- TTS details injected at install-time only when needed
- Preserves provider-agnostic design principle
Changes made:
- src/core/workflows/party-mode/instructions.md
- Reverted .bmad to {bmad_folder} variable
- Replaced hardcoded hook guidance with <!-- TTS_INJECTION:party-mode -->
- Removed <note> about play-tts.sh hook location
- tools/cli/installers/lib/core/installer.js
- Added hook enforcement to party-mode injection replacement
- Guidance now injected only when enableAgentVibes is true
Addresses review comments from bmadcode:
- "needs to remain the variable. it will get set in the file at the install destination."
- "items like this we will need to inject if user is using claude and TTS"
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
* fix: Change 'claude-code' to 'claude' in test script instructions
The correct command to start Claude is 'claude', not 'claude-code'.
Updated line 362-363 in test-bmad-pr.sh to show the correct command.
* fix: Remove npm link from test script to avoid global namespace pollution
- Removed 'npm link' command that was installing BMAD globally
- Changed 'bmad install' to direct node execution using local clone
- Updated success message to reflect no global installation
This keeps testing fully isolated and prevents conflicts with:
- Existing BMAD installations
- Future official BMAD installs
- Orphaned symlinks when test directory is deleted
The test script now runs completely self-contained without modifying
the user's global npm environment.
---------
Co-authored-by: Claude Code <claude@anthropic.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Paul Preibisch <paul@paulpreibisch.com>
Co-authored-by: Brian <bmadcode@gmail.com>
- Add deployment-aware handler generation (filters web-only/ide-only commands)
- Remove unused run-workflow handler type (ghost handler cleanup)
- Implement missing validate-workflow and data handler generation
- Update schema validation to support exactly 6 active handler types
- Clean up activation templates and web bundler logic
- Prevent generation of unused handler instructions for better performance
- All 62 tests pass with backward compatibility maintained
MAJOR BREAKING CHANGES: Phase 4 completely reengineered for developer efficiency and quality
🚀 **Phase 4 Streamlined & Supercharged:**
- **Reduced from 11 to 5 essential workflows** (55% reduction in complexity)
- **Eliminated redundant steps** that created token waste and confusion
- **Created single source of truth** story files with comprehensive implementation context
- **Achieved more reliable results** with fewer steps and better developer guidance
💡 **Revolutionary Dev Agent Behavior Fixes:**
- **Story file is now LAW:** Tasks/subtasks sequence is absolutely binding
- **Red-green-refactor enforcement:** Tests written first, validated, then implementation
- **Zero tolerance for cheating:** Tests must ACTUALLY exist and pass before marking complete
- **Sequential execution only:** No more "doing whatever you want" - follow the story exactly
- **Continuous execution:** No premature pausing until all tasks complete
🎯 **Quality Competition System:**
- **Enhanced story context engine** prevents common LLM development mistakes
- **Quality competition between LLMs** ensures optimal story preparation
- **Comprehensive anti-pattern prevention** stops wheel reinvention and wrong approaches
- **Developer optimization focus** for maximum clarity with minimum verbosity
📋 **Enhanced Definition of Done:**
- **27-point validation checklist** covers all implementation aspects
- **Multiple validation gates** prevent claiming work that isn't actually done
- **Comprehensive test requirements** ensure no functionality goes untested
- **File tracking and documentation** for complete project visibility
🔧 **Technical Improvements:**
- **Variable consistency** throughout all workflow files
- **XML instruction format** for better workflow engine compatibility
- **Proper ask tag handling** for user interaction clarity
- **Project context integration** without blocking implementation
- **Fixed all agent schema compliance** for proper array formatting
**Result:** Phase 4 now delivers superior development outcomes with:
- ✅ **55% fewer workflows** to learn and maintain
- ✅ **Dramatically reduced token usage** and context switching
- ✅ **Eliminated dev agent behavioral issues** that caused quality problems
- ✅ **Faster time-to-completion** with more reliable, predictable results
- ✅ **Better developer experience** with clearer guidance and validation
This represents the most significant Phase 4 improvement since BMAD Method inception - fundamentally fixing developer workflow quality while drastically simplifying the implementation process.
Major Changes:
- Add sample custom agents demonstrating installable agent system
- commit-poet: Generates semantic commit messages (BMAD Method repo sample)
- toolsmith: Development tooling expert with knowledge base covering bundlers, deployment, docs, installers, modules, and tests (BMAD Method repo sample)
- Both agents demonstrate custom agent architecture and are installable to projects via BMAD installer system
- Include comprehensive installation guides and sidecar knowledge bases
- Add bmad-quick-flow methodology for rapid development
- create-tech-spec: Direct technical specification workflow
- quick-dev: Flexible execution workflow supporting both tech-spec-driven and direct instruction development
- quick-flow-solo-dev (Barry): 1 man show agent specialized in bmad-quick-flow methodology
- Comprehensive documentation for quick-flow approach and solo development
- Remove deprecated tech-spec workflow track
- Delete entire tech-spec workflow directory and templates
- Remove quick-spec-flow.md documentation (replaced by quick-flow docs)
- Clean up unused epic and story templates
- Fix custom agent installation across IDE installers
- Repair antigravity and multiple IDE installers to properly support custom agents
- Enable custom agent installation via quick installer, agent installer, regular installer, and special agent installer
- All installation methods now accessible via npx with full documentation
Infrastructure:
- Update BMM module configurations and team setups
- Modify workflow status paths to support quick-flow integration
- Reorganize documentation with new agent and workflow guides
- Add custom/ directory for user customizations
- Update platform codes and installer configurations
## Problem
Custom agents showed generic names (like "Commit Poet") instead of their
actual persona names (like "Inkwell Von Comitizen") in the agent manifest.
## Root Cause
The extractManifestData function was using metadata.name/title instead of
extracting the persona name from the compiled agent XML.
## Solution
1. Added extractAgentAttribute function to pull attributes from <agent> tag
2. Prioritize XML extraction over metadata for persona info:
- displayName: uses agent title attribute from XML
- title: uses agent title attribute from XML
- icon: uses agent icon attribute from XML
- Falls back to metadata if XML extraction fails
## Result
Custom agents now display their actual persona names in manifests:
- Before: "Commit Poet"
- After: "Inkwell Von Comitizen"
This provides better user experience with proper agent identification
in IDE integrations and manifests.
## Problem
Custom agents were only installing to Claude Code (.claude/commands/)
but not to Antigravity (.agent/) or other IDEs that lack installCustomAgentLauncher function.
## Root Cause
Antigravity was missing the installCustomAgentLauncher function that the
IdeManager calls to install custom agents during agent installation.
## Solution
Added installCustomAgentLauncher function to Antigravity that:
- Creates .agent directory if needed
- Generates custom agent launchers with @agentPath references
- Uses same pattern as existing Antigravity agent launchers
- Returns proper installation result for tracking
## Result
Custom agents now install to:
- Claude Code: .claude/commands/bmad/custom/agents/ ✅
- Antigravity: .agent/bmad-custom-agents-{agentName}.md ✅
- Codex: (already working) ✅
All configured IDEs now receive custom agent installations!
## Problem
Compile Agents ignored custom agents in source locations like:
- {project-root}/custom/src/agents/
- {bmad-folder}/custom/src/agents/
## Solution
Update reinstallCustomAgents to check all locations:
1. _cfg/custom/agents/ (backup location)
2. {bmad-folder}/custom/src/agents/ (source in BMAD folder)
3. {project-root}/custom/src/agents/ (source at project root)
## Changes
- Search multiple locations for agents during compile
- Avoid duplicate processing with Set tracking
- Auto-backup source YAML to _cfg/custom/agents/ if needed
- Works with any custom bmad_folder name from config
Now 'Compile Agents' works with agents in any source location!
- Remove verbose explanations and marketing language
- Keep only essential commands and process steps
- Reduce from verbose guide to concise reference
- Focus on what users need to know, not explanations
- Change --path → --source (much clearer purpose)
- Change --target → --destination (more intuitive)
- Update all code references and documentation
- Add advanced parameter examples to installation guide
- Keep short aliases: -s for --source, -t for --destination
Parameters are now much more self-explanatory:
- --source: where to find the agent YAML
- --destination: where to install the compiled agent
- Fix npx bmad → npx bmad-method in documentation
- Only bmad-method is registered in npm registry, not bmad
- Clarify that bmad command works locally when BMAD is installed
- Update installation guides to use correct package name
Ensures users don't get 'package not found' errors when trying npx bmad
- Add npx bmad-method agent-install option for users without cloned repo
- Show both local and npx commands side by side
- Clarify when to use each installation option
- Update automatic updates section with npx option
- Add prominent note about npx capability
Users can now install custom agents without needing to clone the BMAD repository.
* fix: enabled web bundles for test and dev
* fix: only bundle non webskip agents
* fix: addressed pr comments
* fix: addressed pr comments
---------
Co-authored-by: Murat Ozcan <murat@mac.lan>
Antigravity respects .gitignore rules which blocks access to workflow files
in .agent/workflows/, preventing custom workflows from being discovered.
Removing .agent from gitignore to allow Antigravity to scan workflow files.
Note: Codex and Claude Code ignore .gitignore when scanning for custom
prompts, so this only affects Antigravity users.
- Add persistent warning loop when no tools selected in installer
- Users must press spacebar to select, not just highlight
- Red warning explains the issue and offers to go back
- Only way to proceed without tools is explicit "No" confirmation
- Promote Google Antigravity to preferred/recommended IDE section
* feat: Add Google Antigravity IDE installer
Implements installer for Google Antigravity IDE with flattened slash command
naming to match Antigravity's namespace requirements.
Key features:
- Flattened file naming (bmad-module-agents-name.md) for proper slash commands
- Subagent installation support (project-level or user-level)
- Module-specific injection configuration
- Agent, workflow, task, and tool command generation
Implementation:
- Added AntigravitySetup class extending BaseIdeSetup
- Extracted flattenFilename() to BaseIdeSetup for reuse across IDE handlers
- Uses .agent/workflows directory structure
- Supports both interactive and non-interactive configuration
Fixes:
- Proper namespace isolation: /bmad-module-agents-dev instead of /dev
- Prevents conflicts between modules with same agent names
Note: This installer shares 95% of its code with claude-code.js.
Future refactoring could extract common patterns to IdeWithSlashCommandsSetup
base class (see design documents for details).
* chore: update gitignore for antigravity installer
---------
Co-authored-by: Brian <bmadcode@gmail.com>
- Comprehensive alpha.11 changelog capturing all major features
- Condensed earlier alpha releases (0-10) to 5 key bullets each
- Reduced changelog from 1,744 to 609 lines for better readability
- Updated README status badges and installation instructions
## The Tale of the Frame Expert
Once upon a time, BMad Method had a specialized agent called Frame Expert.
This agent was the master of all visual artifacts - flowcharts, diagrams,
wireframes, data flows. Whenever anyone needed a diagram, they called upon
Frame Expert. The agent lived in its own isolated domain with four dedicated
workflows and a library of shared templates.
## The Awakening
But something felt wrong. Teams using BMad Method were meant to mirror real
agile teams - Product Managers, Architects, UX Designers, Tech Writers,
Developers. Each agent represented an authentic role you'd find in any
software team.
Except Frame Expert.
No real agile team has a "Frame Expert" or "Diagram Specialist" who creates
all visual artifacts. In real teams, Architects diagram system architecture.
PMs flowchart processes. UX Designers wireframe interfaces. Tech Writers
create documentation diagrams. The visuals emerge from the domain experts
who need them, not from a centralized diagram factory.
Frame Expert was an abstraction that made technical sense but violated the
very soul of BMad Method - authentic agile role modeling.
## The Transformation
And so Frame Expert was dissolved, its knowledge distributed to those who
truly needed it:
**The Architect** inherited system architecture diagrams and data flows -
the blueprints of technical systems they design.
**The Product Manager** received process flowcharts - the visual maps of
features and workflows they orchestrate.
**The UX Designer** claimed wireframes - the interface sketches that bring
their vision to life.
**The Tech Writer** gained all diagram types - the visual aids that clarify
their documentation.
Each agent now creates diagrams in their domain, using their expertise,
serving their purpose.
## The Shared Knowledge
But the wisdom of diagram creation itself - the Excalidraw templates, the
component libraries, the validation patterns - this knowledge was too
valuable to scatter. It was elevated to core resources, where both BMM
agents AND the new CIS presentation-master agent could draw upon it.
Shared infrastructure for common needs. Distributed execution for domain
expertise.
## The Ripple Effects
With diagrams now properly distributed, other misalignments became visible:
Epic creation was happening in Phase 2 (Planning), before Architecture
existed. But epics need architectural context - API contracts, data models,
technical decisions. So epic creation migrated to Phase 3 (Solutioning),
after Architecture provides that foundation.
Workflow paths were updated. Documentation gained visual flowcharts showing
the complete journey. Agent naming standards were clarified - filenames are
stable roles, persona names are user dreams.
## What Changed
**Removed:**
- frame-expert.agent.yaml (the centralized specialist)
- All frame-expert workflows and shared resources
- Phase 2 epic creation workflow (wrong timing)
- game-design workflow path (consolidated to method track)
- v6-open-items.md (planning doc, now complete)
**Distributed Diagram Capabilities:**
- Architect: create-excalidraw-diagram, create-excalidraw-dataflow
- PM: create-excalidraw-flowchart
- Tech Writer: create-excalidraw-{diagram,dataflow,flowchart}, generate-mermaid
- UX Designer: create-excalidraw-wireframe
**Created:**
- src/core/resources/ (shared diagram context for all modules)
- src/modules/cis/agents/presentation-master.agent.yaml (visual comms specialist)
- src/modules/bmm/workflows/3-solutioning/create-epics-and-stories/ (epic creation's new home)
- src/modules/bmm/workflows/diagrams/ (distributed diagram implementations)
- src/modules/bmm/docs/images/ (workflow visualization assets)
**Enhanced:**
- All agent definitions with domain-appropriate diagram workflows
- Documentation with embedded workflow diagrams and visual guides
- Agent compilation docs with critical naming convention rules
- All 4 workflow paths (enterprise/method × brownfield/greenfield)
**Fixed:**
- Epic creation now in Phase 3 after Architecture
- Story context path variables in BMGD module
- PRD workflow descriptions (epics moved to Phase 3)
## For Users
The Frame Expert commands are gone. In their place:
- Need architecture diagrams? Ask `/architect`
- Need process flows? Ask `/pm`
- Need wireframes? Ask `/ux-designer`
- Need documentation visuals? Ask `/tech-writer`
Each expert creates diagrams in their domain, with their context, using
their judgment.
This is how real teams work.
## Overview
Major enhancement to edit-agent workflow to match create-agent quality standards, plus critical addition of Expert agent sidecar file support and consolidation of validation checklists into single source of truth.
## 1. edit-agent Workflow Comprehensive Enhancement
### Documentation Reference Updates (workflow.yaml)
**Fixed all broken references** - replaced deleted docs with new comprehensive guides:
- ❌ Removed: agent-types.md, agent-architecture.md, agent-command-patterns.md, communication-styles.md
- ✅ Added structured references:
* Core Concepts: understanding_agent_types, agent_compilation
* Architecture Guides: simple_architecture, expert_architecture, module_architecture
* Design Patterns: menu_patterns, communication_presets, brainstorm_context
* Reference Agents: commit-poet, journal-keeper, module examples, BMM agents
### Critical Persona Field Guidance Added (instructions.md +59 lines)
**The #1 issue in legacy agents** - comprehensive guidance on persona field separation:
- Explains how LLMs interpret each field:
* role → "What knowledge/skills/capabilities do I possess?"
* identity → "What background/experience/context shapes my responses?"
* communication_style → "What verbal patterns/word choice do I use?"
* principles → "What beliefs/philosophy drive my choices?"
- BEFORE/AFTER examples showing common mistakes
- Red flag word detection guide (ensures, experienced, believes in, etc.)
- Pure communication style examples from reference agents
### Enhanced Step 1: Analysis (instructions.md +57 lines)
- References all new comprehensive documentation
- **CRITICAL: Persona field separation analysis**
* Checks for behaviors/role/identity mixed into communication_style
* Compares against communication_presets for purity
* Compares against reference agents for quality
- Warm, conversational feedback explaining issues found
### Massive Step 3 Enhancement: Communication Style Refinement (+122 lines)
**7-step prescriptive pattern for fixing the #1 quality issue:**
1. Diagnose Current Communication Style - red flag word detection
2. Extract Non-Style Content - working copy methodology
3. Discover TRUE Communication Style - interview questions + preset exploration
4. Craft Pure Communication Style - good/bad examples from references
5. Show Before/After With Full Context - complete transformation
6. Validate Against Standards - zero red flags, compare to presets/references
7. Confirm With User - explain changes, read dramatically, refine
Examples from actual reference agents:
- "Treats analysis like a treasure hunt - excited by every clue" (Mary/analyst)
- "Ultra-succinct. Speaks in file paths and AC IDs" (Amelia/dev)
- "Poetic drama and flair with every turn of a phrase" (commit-poet)
### Legacy Type Migration Pattern (+42 lines)
**Comprehensive guide for full/hybrid/standalone → Simple/Expert/Module migration:**
- Clear explanations of modern types (architecture, NOT capability)
- Migration patterns with decision tree
- Structural conversion guides (Simple ↔ Expert)
- Module = design intent clarification
### Enhanced Validation (Step 4)
- Persona field separation validation emphasized
- Updated success message with all quality standards
- Comprehensive checklist validation
### Complete README Documentation (200 lines created)
- Purpose emphasizing persona field separation as #1 issue
- 14 common editing scenarios (persona separation listed FIRST)
- Complete doc reference listing by category
- Dedicated "Critical: Persona Field Separation" section
- Red flag words guide
- 3 detailed usage scenarios
- Quality standards checklist
## 2. Expert Agent Sidecar File Support (NEW)
### Step 1: Smart Path Detection and Loading (+35 lines)
**Automatically detects and loads based on path type:**
```yaml
If path is .agent.yaml → Simple Agent (load single file)
If path is folder → Expert Agent:
- Load .agent.yaml from inside folder
- Load ALL sidecar files (*.md, *.txt, *.csv, *.json, *.yaml)
- Create inventory for reference
- Present: "Loaded agent.yaml + 5 sidecar files: [list]"
```
**Sidecar analysis:**
- Maps which menu items reference which sidecar files (tmpl="path", data="path")
- Checks if all sidecar references actually exist
- Identifies unused/orphaned sidecar files
- Assesses sidecar organization
**Warm expert agent feedback:**
"This is beautifully organized as an Expert agent! The sidecar files include 3 journal templates (daily, weekly, breakthrough) and a mood-patterns knowledge file. Your menu items reference them nicely. I do notice 'old-template.md' isn't referenced anywhere - we could clean that up."
### Step 3: Sidecar Editing Patterns (+47 lines)
**5 complete sidecar editing scenarios:**
1. **Updating templates** - Edit content, verify references work, test variables
2. **Adding new sidecar files** - Create file + add menu item with reference
3. **Removing unused sidecar files** - Confirm unused, ask to delete, clean references
4. **Reorganizing sidecar structure** - Move files, update ALL YAML references
5. **Updating knowledge base files** - Edit .csv/.json/.yaml data directly
**Critical mindset:** "Sidecar files are as much a part of the agent as the YAML!"
### Step 4: Sidecar Validation (+16 lines)
**Conversational validation:**
- "Your menu item 'daily-journal' references 'templates/daily.md'... checking... ✓ exists!"
- Check for orphaned files not referenced anywhere
- Verify sidecar file formats (YAML parses, CSV has headers, markdown well-formed)
- Success message: "✓ All sidecar file references valid - 5 sidecar files, all referenced correctly!"
### README Updates
- "What You'll Need" distinguishes Simple vs Expert paths
- Scenario 2b: Complete Expert agent editing example (journal-keeper template update)
- Updated common scenarios list
## 3. Unified Validation Checklist (Single Source of Truth)
### Problem Solved
- create-agent had outdated checklist (62 lines, no persona field separation)
- edit-agent had enhanced checklist (112 lines, with our improvements)
- Risk of drift and inconsistency between workflows
### Solution: agent-validation-checklist.md (160 lines)
**Canonical location:** `/src/modules/bmb/workflows/create-agent/`
**Comprehensive coverage combining best of both:**
- YAML structure validation
- **Persona field separation** (field separation check, purity check, quality benchmarking)
- Menu validation (including sidecar file path validation)
- Type-specific validation (Simple/Expert/Module)
- Compilation validation
- Common issues and fixes section
**Key sections:**
- Persona Validation (CRITICAL - #1 Quality Issue)
* Field Separation Check (what belongs where)
* Communication Style Purity Check (red flag word detection)
* Quality Benchmarking (compare to presets and references)
- Expert Agent validation (9 sidecar-specific checks)
- Module Agent validation (design intent verification)
- Common Issues and Fixes (real examples with solutions)
### References Updated
**create-agent/workflow.yaml:**
```yaml
validation: "{installed_path}/agent-validation-checklist.md"
```
**edit-agent/workflow.yaml:**
```yaml
# Shared validation checklist (canonical location in create-agent folder)
validation: "{project-root}/{bmad_folder}/bmb/workflows/create-agent/agent-validation-checklist.md"
```
**edit-agent/instructions.md:**
```xml
<note>The validation checklist is shared between create-agent and edit-agent workflows to ensure consistent quality standards.</note>
```
### Files Changed
- ✅ Created: agent-validation-checklist.md (160 lines)
- ❌ Deleted: create-agent/checklist.md (62 lines)
- ❌ Deleted: edit-agent/checklist.md (112 lines)
- Updated: Both workflow.yaml files to reference unified checklist
## Statistics
**Overall changes:** 6 files changed, +553 insertions, -264 deletions
**edit-agent enhancements:**
- instructions.md: +396 lines (comprehensive guidance)
- README.md: +213 lines (complete documentation)
- workflow.yaml: +32 lines (proper doc references)
**Validation unification:**
- Net result: Better quality, less duplication, easier maintenance
- Single source of truth for all agent validation
## Impact
### For Users Editing Agents
- Automatically detects and handles Expert agents with sidecar files
- Clear guidance on fixing #1 issue (persona field separation)
- 7-step prescriptive pattern for communication style refinement
- Warm, educational feedback throughout
- Validates against same standards as create-agent
### For Agent Quality
- Both create and edit workflows use same validation standards
- Persona field separation gets proper attention
- Expert agent sidecar files treated as first-class citizens
- Legacy agents can be migrated to modern standards
- All agents validated against reference implementations
### For Maintenance
- ONE checklist to maintain instead of two
- Consistent quality standards across workflows
- Documentation properly linked to new comprehensive guides
- No risk of checklist drift
This brings edit-agent to the same quality level as create-agent while adding critical Expert agent support and establishing single source of truth for validation.
## Overview
This commit represents a complete overhaul of the BMAD agent creation system, establishing clear standards for agent development, installation workflows, and persona design. The changes span documentation, tooling, reference implementations, and field-specific guidance.
## Key Components
### 1. Agent Installation Infrastructure
**New CLI Command: `agent-install`**
- Interactive agent installation with persona customization
- Supports Simple (single YAML), Expert (sidecar files), and Module agents
- Template variable processing with Handlebars-style syntax
- Automatic compilation from YAML to XML (.md) format
- Manifest tracking and IDE integration (Claude Code, Cursor, Windsurf, etc.)
- Source preservation in `_cfg/custom/agents/` for reinstallation
**Files Created:**
- `tools/cli/commands/agent-install.js` - Main CLI command
- `tools/cli/lib/agent/compiler.js` - YAML to XML compilation engine
- `tools/cli/lib/agent/installer.js` - Installation orchestration
- `tools/cli/lib/agent/template-engine.js` - Handlebars template processing
**Compiler Features:**
- Auto-injects frontmatter, activation, handlers, help/exit menu items
- Smart handler inclusion (only includes action/workflow/exec/tmpl handlers actually used)
- Proper XML escaping and formatting
- Persona name customization (e.g., "Fred the Commit Poet")
### 2. Documentation Overhaul
**Deleted Bloated/Outdated Docs (2,651 lines removed):**
- Old verbose architecture docs
- Redundant pattern files
- Outdated workflow guides
**Created Focused, Type-Specific Docs:**
- `src/modules/bmb/docs/understanding-agent-types.md` - Architecture vs capability distinction
- `src/modules/bmb/docs/simple-agent-architecture.md` - Self-contained agents
- `src/modules/bmb/docs/expert-agent-architecture.md` - Agents with sidecar files
- `src/modules/bmb/docs/module-agent-architecture.md` - Workflow-integrated agents
- `src/modules/bmb/docs/agent-compilation.md` - YAML → XML process
- `src/modules/bmb/docs/agent-menu-patterns.md` - Menu design patterns
- `src/modules/bmb/docs/index.md` - Documentation hub
**Net Result:** ~1,930 line reduction while adding MORE value through focused content
### 3. Create-Agent Workflow Enhancements
**Critical Persona Field Guidance Added to Step 4:**
Explains how the LLM interprets each persona field when the agent activates:
- **role** → "What knowledge, skills, and capabilities do I possess?"
- **identity** → "What background, experience, and context shape my responses?"
- **communication_style** → "What verbal patterns, word choice, quirks, and phrasing do I use?"
- **principles** → "What beliefs and operating philosophy drive my choices?"
**Key Insight:** `communication_style` should ONLY describe HOW the agent talks, not restate role/identity/principles. The `communication-presets.csv` provides 60 pure communication styles with NO role/identity/principles mixed in.
**Files Updated:**
- `src/modules/bmb/workflows/create-agent/instructions.md` - Added persona field interpretation guide
- `src/modules/bmb/workflows/create-agent/brainstorm-context.md` - Refined to 137 lines
- `src/modules/bmb/workflows/create-agent/communication-presets.csv` - 60 styles across 13 categories
### 4. Reference Agent Cleanup
**Removed install_config Personality Bloat:**
Understanding: Future installer will handle personality customization, so stripped all personality toggles from reference agents.
**commit-poet.agent.yaml** (Simple Agent):
- BEFORE: 36 personality combinations (3 enthusiasm × 3 depths × 4 styles) = decision fatigue
- AFTER: Single concise persona with pure communication style
- Changed from verbose conditionals to: "Poetic drama and flair with every turn of a phrase. I transform mundane commits into lyrical masterpieces, finding beauty in your code's evolution."
- Reduction: 248 lines → 153 lines (38% reduction)
**journal-keeper.agent.yaml** (Expert Agent):
- Stripped install_config, simplified communication_style
- Shows proper Expert agent structure with sidecar files
**security-engineer.agent.yaml & trend-analyst.agent.yaml** (Module Agents):
- Added header comments explaining WHY Module Agent (design intent, not just location)
- Clarified: Module agents are designed FOR ecosystem integration, not capability-limited
**Files Updated:**
- `src/modules/bmb/reference/agents/simple-examples/commit-poet.agent.yaml`
- `src/modules/bmb/reference/agents/expert-examples/journal-keeper/journal-keeper.agent.yaml`
- `src/modules/bmb/reference/agents/module-examples/security-engineer.agent.yaml`
- `src/modules/bmb/reference/agents/module-examples/trend-analyst.agent.yaml`
### 5. BMM Agent Voice Enhancement
**Gave all 9 BMM agents distinct, memorable communication voices:**
**Mary (analyst)** - The favorite! Changed from generic "systematic and probing" to:
"Treats analysis like a treasure hunt - excited by every clue, thrilled when patterns emerge. Asks questions that spark 'aha!' moments while structuring insights with precision."
**Other Notable Voices:**
- **John (pm):** "Asks 'WHY?' relentlessly like a detective on a case. Direct and data-sharp, cuts through fluff to what actually matters."
- **Winston (architect):** "Speaks in calm, pragmatic tones, balancing 'what could be' with 'what should be.' Champions boring technology that actually works."
- **Amelia (dev):** "Ultra-succinct. Speaks in file paths and AC IDs - every statement citable. No fluff, all precision."
- **Bob (sm):** "Crisp and checklist-driven. Every word has a purpose, every requirement crystal clear. Zero tolerance for ambiguity."
- **Sally (ux-designer):** "Paints pictures with words, telling user stories that make you FEEL the problem. Empathetic advocate with creative storytelling flair."
**Pattern Applied:** Moved behaviors from communication_style to principles, keeping communication_style as PURE verbal patterns.
**Files Updated:**
- `src/modules/bmm/agents/analyst.agent.yaml`
- `src/modules/bmm/agents/pm.agent.yaml`
- `src/modules/bmm/agents/architect.agent.yaml`
- `src/modules/bmm/agents/dev.agent.yaml`
- `src/modules/bmm/agents/sm.agent.yaml`
- `src/modules/bmm/agents/tea.agent.yaml`
- `src/modules/bmm/agents/tech-writer.agent.yaml`
- `src/modules/bmm/agents/ux-designer.agent.yaml`
- `src/modules/bmm/agents/frame-expert.agent.yaml`
### 6. Linting Fixes
**ESLint Compliance:**
- Replaced all `'utf-8'` with `'utf8'` (unicorn/text-encoding-identifier-case)
- Changed `variables.hasOwnProperty(varName)` to `Object.hasOwn(variables, varName)` (unicorn/prefer-object-has-own)
- Replaced `JSON.parse(JSON.stringify(...))` with `structuredClone(...)` (unicorn/prefer-structured-clone)
- Fixed empty YAML mapping values in sample files
**Files Fixed:**
- 7 JavaScript files across agent tooling (compiler, installer, commands, IDE integration)
- 1 YAML sample file
## Architecture Decisions
### Agent Types Are About Architecture, Not Capability
- **Simple:** Self-contained in single YAML (NOT limited in capability)
- **Expert:** Includes sidecar files (templates, docs, etc.)
- **Module:** Designed for BMAD ecosystem integration (workflows, cross-agent coordination)
### Persona Field Separation Critical for LLM Interpretation
The LLM needs distinct fields to understand its role:
- Mixing role/identity/principles into communication_style confuses the persona
- Pure communication styles (from communication-presets.csv) have ZERO role/identity/principles content
- Example DON'T: "Experienced analyst who uses systematic approaches..." (mixing identity + style)
- Example DO: "Systematic and probing. Structures findings hierarchically." (pure style)
### Install-Time vs Runtime Configuration
- Template variables ({{var}}) resolve at compile-time
- Runtime variables ({user_name}, {bmad_folder}) resolve when agent activates
- Future installer will handle personality customization, so agents should ship with single default persona
## Testing
- All linting passes (ESLint with max-warnings=0)
- Agent compilation tested with commit-poet, journal-keeper examples
- Install workflow validated with Simple and Expert agent types
- Manifest tracking and IDE integration verified
## Impact
This establishes BMAD as having a complete, production-ready agent creation and installation system with:
- Clear documentation for all agent types
- Automated compilation and installation
- Strong persona design guidance
- Reference implementations showing best practices
- Distinct, memorable agent voices throughout BMM module
Co-Authored-By: BMad Builder <builder@bmad.dev>
Co-Authored-By: Mary the Analyst <analyst@bmad.dev>
Co-Authored-By: Paige the Tech Writer <tech-writer@bmad.dev>
Implements installer for Google Antigravity IDE with flattened slash command
naming to match Antigravity's namespace requirements.
Key features:
- Flattened file naming (bmad-module-agents-name.md) for proper slash commands
- Subagent installation support (project-level or user-level)
- Module-specific injection configuration
- Agent, workflow, task, and tool command generation
Implementation:
- Added AntigravitySetup class extending BaseIdeSetup
- Extracted flattenFilename() to BaseIdeSetup for reuse across IDE handlers
- Uses .agent/workflows directory structure
- Supports both interactive and non-interactive configuration
Fixes:
- Proper namespace isolation: /bmad-module-agents-dev instead of /dev
- Prevents conflicts between modules with same agent names
Note: This installer shares 95% of its code with claude-code.js.
Future refactoring could extract common patterns to IdeWithSlashCommandsSetup
base class (see design documents for details).
Add support for choosing between global and project-specific installation
locations for Codex CLI prompts with CODEX_HOME configuration instructions.
Changes:
- Add collectConfiguration() to prompt for installation location (default: global)
- Support global (~/.codex/prompts) and project-specific (<project>/.codex/prompts)
- Display OS-specific CODEX_HOME setup instructions before user confirms
- Update detect() and cleanup() to handle both installation locations
Installation options:
- Global: Simple, works immediately, but prompts reference specific .bmad path
- Project-specific: Requires CODEX_HOME, better for multi-project workflows
- Unix/Mac: alias codex='CODEX_HOME="$PWD/.codex" codex'
- Windows: codex.cmd wrapper with %~dp0
Major milestone: Epics & Stories now generated AFTER Architecture
- New Frame Expert agent with Excalidraw workflows
- Time estimate prohibition across all workflows
- Platform-specific command filtering (ide-only/web-only)
- Agent customization enhancement (prompts & memories)
- Workflow configuration standardization
- Add merging logic for customizeYaml.prompts and customizeYaml.memories in loadAndMergeAgent()
- Implement buildMemoriesXml() method to output memories XML section
- Update buildPromptsXml() to use <content> wrapper instead of CDATA for better compatibility
- Integrate memories output into convertToXml() between persona and prompts sections
Changes:
1. Line 123-131: Added prompts and memories append logic in loadAndMergeAgent()
2. Line 215-218: Added memories XML output in convertToXml()
3. Line 286-301: New buildMemoriesXml() method
4. Line 312-315: Updated prompts to use <content> wrapper for consistency
This allows users to customize agents via bmad/_cfg/agents/*.customize.yaml with:
- prompts: Array of {id, content} objects for action handlers
- memories: Array of strings for persistent agent memories
## Summary
- Add ide-only and web-only boolean fields to agent menu schema
- Filter menu items based on build target (web bundle vs local IDE)
- Update BMM agent definitions with platform restrictions and improved descriptions
- Update frame-expert agent icon to 📐 and add webskip flag
## Changes
### Schema & Bundler Updates
- tools/schema/agent.js: Add ide-only and web-only optional boolean fields
- tools/cli/lib/yaml-xml-builder.js: Filter ide-only items from web bundles
- tools/cli/lib/xml-handler.js: Pass forWebBundle flag through build chain
### Agent Updates
- frame-expert: Change icon to 📐, add webskip flag, improve principle formatting
- pm: Mark workflow-init and correct-course as ide-only, advanced-elicitation as web-only
- ux-designer: Rename trigger to create-ux-design, mark advanced-elicitation as web-only
- sm: Rename triggers for consistency (story-context → create-story-context)
- analyst: Add research workflow after brainstorm, mark advanced-elicitation as web-only
- architect: Remove document field from validate-architecture, add web-only flag
- dev: Update persona and critical actions for clarity
- tech-writer: Add party-mode workflow, mark advanced-elicitation as web-only
- tea: Mark advanced-elicitation as web-only
- All agents: Standardize party-mode description
This enables platform-specific functionality where some commands only make sense
in IDE environments (workflow-init) or web interfaces (advanced-elicitation).
