Brian Madison 48cf5c8056 feat(workflows): Implement intelligent file discovery protocol and Phase 4 BMGD workflows
## Core Workflow Engine Enhancements

### discover_inputs Protocol (MAJOR)
- Added reusable `discover_inputs` protocol to workflow.xml for intelligent file loading
- Supports three loading strategies:
  - FULL_LOAD: Load all shards for PRD, Architecture, UX (changed pattern from /index.md to /*/*.md)
  - SELECTIVE_LOAD: Load specific shard via template variable (e.g., epic-{{epic_num}}.md)
  - INDEX_GUIDED: Load index, analyze TOC, intelligently load relevant docs (with "DO NOT BE LAZY" mandate)
- Auto-discovers whole vs sharded documents with proper fallback
- Provides transparent reporting of loaded content with file counts
- Invoked via <invoke-protocol name="discover_inputs" /> tag in workflow instructions

### Advanced Elicitation Improvements
- Renamed adv-elicit.xml to advanced-elicitation.xml for clarity
- Updated all references across agents and commands

### Shard Document Tool Enhancement
- Added Step 6: Handle Original Document with three options:
  - [d] Delete - Remove original (recommended, prevents confusion)
  - [m] Move to archive - Backup original to archive folder
  - [k] Keep - Warning about defeating sharding purpose
- Prevents issue where both whole and sharded versions exist, confusing discover_inputs protocol

## BMM Module - Input File Pattern Standardization

### Phase 1 - Analysis (1 workflow)
- product-brief: Added load_strategy (FULL_LOAD for research/brainstorming, INDEX_GUIDED for document_project)
- Updated instructions.md to use invoke-protocol, replaced manual fuzzy matching

### Phase 2 - Planning (4 workflows)
- prd: Added load_strategy, updated instructions to reference {product_brief_content}, {research_content}
- create-ux-design: Added load_strategy, removed fuzzy matching from instructions
- tech-spec: Added load_strategy for brownfield context discovery
- All epics patterns updated to support SELECTIVE_LOAD for specific epic shards

### Phase 3 - Solutioning (2 workflows)
- architecture: Added load_strategy, updated instructions to use pre-loaded {prd_content}, {epics_content}, {ux_design_content}
- solutioning-gate-check: Added load_strategy, replaced manual discovery with protocol invocation

### Phase 4 - Implementation (8 workflows)
- code-review: Added load_strategy, fixed sharded patterns to /*/*.md, added step 1.5 for protocol
- correct-course: Added complete input_file_patterns section (was missing), added step 0.5
- create-story: Added load_strategy, updated to SELECTIVE_LOAD for epics, added step 1.5
- dev-story: Added complete input_file_patterns section (was missing), added step 0.5
- epic-tech-context: Added load_strategy, updated PRD extraction to use {prd_content}, added step 1.5
- retrospective: Added load_strategy for architecture/prd (FULL_LOAD), epics (SELECTIVE_LOAD), added step 0.5
- sprint-planning: Fixed sharded pattern to load ALL epics (/*/*.md), added step 0.5
- story-context: Added load_strategy, updated doc collection to reference pre-loaded content, added step 1.5

### Sprint Artifacts Path Corrections
- story-done: Added missing sprint_artifacts variable, fixed sprint_status path from {context_dir} to {sprint_artifacts}
- story-ready: Added missing sprint_artifacts variable
- story-context: Fixed undefined {context_dir} -> {sprint_artifacts}
- correct-course: Added sprint_artifacts and sprint_status variables

## BMGD Module - Phase 4 Production Workflows (NEW)

Added complete Phase 4 implementation workflows for game development:
- code-review: Senior developer review for completed game features
- correct-course: Sprint change management for game projects
- create-story: Story generation for game mechanics/features
- dev-story: Feature implementation workflow
- epic-tech-context: Technical spec generation per game epic
- retrospective: Epic completion review and lessons learned
- sprint-planning: Game development sprint status tracking
- story-context: Dynamic context assembly for game stories
- story-done: Story completion workflow
- story-ready: Story readiness workflow

All BMGD workflows follow BMM patterns with game-specific adaptations.

## Agent Updates

### BMM Agents
- Updated all 7 BMM agents (analyst, architect, dev, pm, sm, tea, tech-writer, ux-designer)
- Standardized web bundle configurations

### BMGD Agents
- Updated 4 game development agents (game-architect, game-designer, game-dev, game-scrum-master)
- Aligned with BMM agent structure

### CIS Agents
- Updated 5 creative intelligence agents for consistency

## Documentation & Configuration

- Updated CHANGELOG.md with Phase 4 workflow additions
- Updated files-manifest.csv and task-manifest.csv
- Updated .claude commands for all agents
- Fixed formatting issues from previous commits

## Breaking Changes

NONE - All changes are backward compatible. Workflows without input_file_patterns continue to work.
Workflows with input_file_patterns now benefit from intelligent auto-loading.

## Migration Notes

Existing workflows can gradually adopt discover_inputs protocol by:
1. Adding load_strategy to existing input_file_patterns in workflow.yaml
2. Adding <invoke-protocol name="discover_inputs" /> step in instructions.md
3. Replacing manual file loading with references to {pattern_name_content} variables
2025-11-12 19:18:38 -06:00

377 lines
14 KiB
Markdown

# Convert Legacy - v4 to v6 Conversion Instructions
<critical>The workflow execution engine is governed by: {project-root}/.bmad/core/tasks/workflow.xml</critical>
<parameter name="You MUST have already loaded and processed: {project-root}/.bmad/bmb/workflows/convert-legacy/workflow.yaml</critical>
<critical>Communicate in {communication_language} throughout the conversion process</critical>
<workflow>
<step n="1" goal="Identify and Load Legacy Item">
<action>Ask user for the path to the v4 item to convert (agent, workflow, or module)</action>
<action>Load the complete file/folder structure</action>
<action>Detect item type based on structure and content patterns:</action>
- Agent: Contains agent or prompt XML tags, single file
- Workflow: Contains workflow YAML or instruction patterns, usually folder
- Module: Contains multiple agents/workflows in organized structure
- Task: Contains task XML tags
<ask>Confirm detected type or allow user to correct: "Detected as [type]. Is this correct? (y/n)"</ask>
</step>
<step n="2" goal="Analyze Legacy Structure">
<action>Parse the v4 structure and extract key components:</action>
For v4 Agents (YAML-based in markdown):
- Agent metadata (name, id, title, icon, whenToUse)
- Persona block (role, style, identity, focus, core_principles)
- Commands list with task/template references
- Dependencies (tasks, templates, checklists, data files)
- Activation instructions and workflow rules
- IDE file resolution patterns
For v4 Templates (YAML-based document generators):
- Template metadata (id, name, version, output)
- Workflow mode and elicitation settings
- Sections hierarchy with:
- Instructions for content generation
- Elicit flags for user interaction
- Templates with {{variables}}
- Conditional sections
- Repeatable sections
For v4 Tasks (Markdown with execution instructions):
- Critical execution notices
- Step-by-step workflows
- Elicitation requirements (1-9 menu format)
- Processing flows and decision trees
- Agent permission rules
For Modules:
- Module metadata
- Component list (agents, workflows, tasks)
- Dependencies
- Installation requirements
<action>Create a conversion map of what needs to be transformed</action>
<action>Map v4 patterns to v6 equivalents:
- v4 Task + Template → v6 Workflow (folder with workflow.yaml, instructions.md, template.md)
- v4 Agent YAML → v6 Agent YAML format
- v4 Commands → v6 <menu> with proper handlers
- v4 Dependencies → v6 workflow references or data files
</action>
</step>
<step n="3" goal="Determine Target Module and Location">
<ask>Which module should this belong to? (eg. bmm, bmb, cis, bmm-legacy, or custom)</ask>
<action if="custom module"><ask>Enter custom module code (kebab-case):</ask></action>
<action>Determine installation path based on type and module</action>
<critical>IMPORTANT: All paths must use final BMAD installation locations, not src paths!</critical>
<action>Show user the target location: {project-root}/.bmad/{{target_module}}/{{item_type}}/{{item_name}}</action>
<action>Note: Files will be created in .bmad/ but all internal paths will reference {project-root}/.bmad/ locations</action>
<ask>Proceed with this location? (y/n)</ask>
</step>
<step n="4" goal="Choose Conversion Strategy">
<action>Based on item type and complexity, choose approach:</action>
<check if="agent conversion">
<check if="simple agent (basic persona + commands)">
<action>Use direct conversion to v6 agent YAML format</action>
<goto step="5a">Direct Agent Conversion</goto>
</check>
<check if="complex agent with embedded workflows">
<action>Plan to invoke create-agent workflow</action>
<goto step="5b">Workflow-Assisted Agent Creation</goto>
</check>
</check>
<check if="template or task conversion to workflow">
<action>Analyze the v4 item to determine workflow type:</action>
- Does it generate a specific document type? → Document workflow
- Does it produce structured output files? → Document workflow
- Does it perform actions without output? → Action workflow
- Does it coordinate other tasks? → Meta-workflow
- Does it guide user interaction? → Interactive workflow
<ask>Based on analysis, this appears to be a {{detected_workflow_type}} workflow. Confirm or correct:
1. Document workflow (generates documents with template)
2. Action workflow (performs actions, no template)
3. Interactive workflow (guided session)
4. Meta-workflow (coordinates other workflows)
Select 1-4:</ask>
<action if="template conversion"><goto step="5c">Template-to-Workflow Conversion</goto></action>
<action if="task conversion"><goto step="5e">Task-to-Workflow Conversion</goto></action>
</check>
<check if="full module conversion">
<action>Plan to invoke create-module workflow</action>
<goto step="5d">Module Creation</goto>
</check>
</step>
<step n="5a" goal="Direct Agent Conversion" optional="true">
<action>Transform v4 YAML agent to v6 YAML format:</action>
1. Convert agent metadata structure:
- v4 `agent.name` → v6 `agent.metadata.name`
- v4 `agent.id` → v6 `agent.metadata.id`
- v4 `agent.title` → v6 `agent.metadata.title`
- v4 `agent.icon` → v6 `agent.metadata.icon`
- Add v6 `agent.metadata.module` field
2. Transform persona structure:
- v4 `persona.role` → v6 `agent.persona.role` (keep as YAML string)
- v4 `persona.style` → v6 `agent.persona.communication_style`
- v4 `persona.identity` → v6 `agent.persona.identity`
- v4 `persona.core_principles` → v6 `agent.persona.principles` (as array)
3. Convert commands to menu:
- v4 `commands:` list → v6 `agent.menu:` array
- Each command becomes menu item with:
- `trigger:` (without \* prefix - added at build)
- `description:`
- Handler attributes (`workflow:`, `exec:`, `action:`, etc.)
- Map task references to workflow paths
- Map template references to workflow invocations
4. Add v6-specific sections (in YAML):
- `agent.prompts:` array for inline prompts (if using action: "#id")
- `agent.critical_actions:` array for startup requirements
- `agent.activation_rules:` for universal agent rules
5. Handle dependencies and paths:
- Convert task dependencies to workflow references
- Map template dependencies to v6 workflows
- Preserve checklist and data file references
- CRITICAL: All paths must use {project-root}/.bmad/{{module}}/ NOT src/
<action>Generate the converted v6 agent YAML file (.agent.yaml)</action>
<action>Example path conversions:
- exec="{project-root}/.bmad/{{target_module}}/tasks/task-name.md"
- run-workflow="{project-root}/.bmad/{{target_module}}/workflows/workflow-name/workflow.yaml"
- data="{project-root}/.bmad/{{target_module}}/data/data-file.yaml"
</action>
<action>Save to: .bmad/{{target_module}}/agents/{{agent_name}}.agent.yaml (physical location)</action>
<action>Note: The build process will later compile this to .md with XML format</action>
<goto step="6">Continue to Validation</goto>
</step>
<step n="5b" goal="Workflow-Assisted Agent Creation" optional="true">
<action>Extract key information from v4 agent:</action>
- Name and purpose
- Commands and functionality
- Persona traits
- Any special behaviors
<invoke-workflow>
workflow: {project-root}/.bmad/bmb/workflows/create-agent/workflow.yaml
inputs:
- agent_name: {{extracted_name}}
- agent_purpose: {{extracted_purpose}}
- commands: {{extracted_commands}}
- persona: {{extracted_persona}}
</invoke-workflow>
<goto step="6">Continue to Validation</goto>
</step>
<step n="5c" goal="Template-to-Workflow Conversion" optional="true">
<action>Convert v4 Template (YAML) to v6 Workflow:</action>
1. Extract template metadata:
- Template id, name, version → workflow.yaml name/description
- Output settings → default_output_file
- Workflow mode (interactive/yolo) → workflow settings
2. Convert template sections to instructions.md:
- Each YAML section → workflow step
- Conditional sections → `if="condition"` attribute
- Repeatable sections → `repeat="for-each"` attribute
- Section instructions → step content
3. Extract template structure to template.md:
- Section content fields → template structure
- {{variables}} → preserve as-is
- Nested sections → hierarchical markdown
4. Handle v4 create-doc.md task integration:
- Elicitation methods (1-9 menu) → convert to v6 elicitation
- Agent permissions → note in instructions
- Processing flow → integrate into workflow steps
<critical>When invoking create-workflow, the standard config block will be automatically added:</critical>
```yaml
# Critical variables from config
config_source: '{project-root}/.bmad/{{target_module}}/config.yaml'
output_folder: '{config_source}:output_folder'
user_name: '{config_source}:user_name'
communication_language: '{config_source}:communication_language'
date: system-generated
```
<invoke-workflow>
workflow: {project-root}/.bmad/bmb/workflows/create-workflow/workflow.yaml
inputs:
- workflow_name: {{template_name}}
- workflow_type: document
- template_structure: {{extracted_template}}
- instructions: {{converted_sections}}
</invoke-workflow>
<action>Verify the created workflow.yaml includes standard config block</action>
<action>Update converted instructions to use config variables where appropriate</action>
<goto step="6">Continue to Validation</goto>
</step>
<step n="5d" goal="Module Creation" optional="true">
<action>Analyze module structure and components</action>
<action>Create module blueprint with all components</action>
<invoke-workflow>
workflow: {project-root}/.bmad/bmb/workflows/create-module/workflow.yaml
inputs:
- module_name: {{module_name}}
- components: {{component_list}}
</invoke-workflow>
<goto step="6">Continue to Validation</goto>
</step>
<step n="5e" goal="Task-to-Workflow Conversion" optional="true">
<action>Convert v4 Task (Markdown) to v6 Workflow:</action>
1. Analyze task purpose and output:
- Does it generate documents? → Create template.md
- Does it process data? → Action workflow
- Does it guide user interaction? → Interactive workflow
- Check for file outputs, templates, or document generation
2. Extract task components:
- Execution notices and critical rules → workflow.yaml metadata
- Step-by-step instructions → instructions.md steps
- Decision trees and branching → flow control tags
- User interaction patterns → appropriate v6 tags
3. Based on confirmed workflow type:
<check if="Document workflow">
- Create template.md from output patterns
- Map generation steps to instructions
- Add template-output tags for sections
</check>
<check if="Action workflow">
- Set template: false in workflow.yaml
- Focus on action sequences in instructions
- Preserve execution logic
</check>
4. Handle special v4 patterns:
- 1-9 elicitation menus → v6 <invoke-task halt="true">{project-root}/.bmad/core/tasks/advanced-elicitation.xml</invoke-task>
- Agent permissions → note in instructions
- YOLO mode → autonomous flag or optional steps
- Critical notices → workflow.yaml comments
<critical>When invoking create-workflow, the standard config block will be automatically added:</critical>
```yaml
# Critical variables from config
config_source: '{project-root}/.bmad/{{target_module}}/config.yaml'
output_folder: '{config_source}:output_folder'
user_name: '{config_source}:user_name'
communication_language: '{config_source}:communication_language'
date: system-generated
```
<invoke-workflow>
workflow: {project-root}/.bmad/bmb/workflows/create-workflow/workflow.yaml
inputs:
- workflow_name: {{task_name}}
- workflow_type: {{confirmed_workflow_type}}
- instructions: {{extracted_task_logic}}
- template: {{generated_template_if_document}}
</invoke-workflow>
<action>Verify the created workflow.yaml includes standard config block</action>
<action>Update converted instructions to use config variables where appropriate</action>
<goto step="6">Continue to Validation</goto>
</step>
<step n="6" goal="Validate Conversion">
<action>Run validation checks on converted item:</action>
For Agents:
- [ ] Valid YAML structure (.agent.yaml)
- [ ] All required sections present (metadata, persona, menu)
- [ ] Menu items properly formatted (trigger, description, handlers)
- [ ] Paths use {project-root} variables
For Workflows:
- [ ] Valid YAML syntax
- [ ] Instructions follow v6 conventions
- [ ] Template variables match
- [ ] File structure correct
**Standard Config Validation (Workflows):**
- [ ] workflow.yaml contains standard config block:
- config_source defined
- output_folder, user_name, communication_language pulled from config
- date set to system-generated
- [ ] Converted instructions use config variables where appropriate
- [ ] Template includes config variables in metadata (if document workflow)
- [ ] No hardcoded paths that should use {output_folder}
- [ ] No generic greetings that should use {user_name}
For Modules:
- [ ] All components converted
- [ ] Proper folder structure
- [ ] Config files valid
- [ ] Installation ready
<action>Show validation results to user</action>
<ask>Any issues to fix before finalizing? (y/n)</ask>
<check if="yes">
<action>Address specific issues</action>
<goto step="6">Re-validate</goto>
</check>
</step>
<step n="7" goal="Migration Report">
<action>Generate conversion report showing:</action>
- Original v4 location
- New v6 location
- Items converted
- Any manual adjustments needed
- Warnings or notes
<action>Save report to: {output_folder}/conversion-report-{{date}}.md</action>
<action>Inform {user_name} in {communication_language} that the conversion report has been generated</action>
</step>
<step n="8" goal="Cleanup and Finalize">
<ask>Archive original v4 files? (y/n)</ask>
<action if="yes">Move v4 files to: {project-root}/archive/v4-legacy/{{date}}/</action>
<action>Show user the final converted items and their locations</action>
<action>Provide any post-conversion instructions or recommendations</action>
<ask>Would you like to convert another legacy item? (y/n)</ask>
<action if="yes"><goto step="1">Start new conversion</goto></action>
</step>
</workflow>