Combined Option A + B implementation providing comprehensive large context analysis:
Core Integration:
- Add gemini-analysis.md task for massive context window operations
- Configure geminiCli settings in core-config.yaml with safety controls
- Support @file/directory syntax and --all-files mode for project-wide analysis
- Implement configurable approval modes and timeout protections
Agent Capabilities:
- Add *gemini-analyze command to analyst, architect, and dev agents
- Enable large codebase discovery, architecture analysis, and debugging
- Support feature verification and pattern discovery across entire projects
- Maintain read-only default with approval prompts for modifications
Team Configurations:
- Create team-gemini-analysis.yaml for specialized large context work
- Include comprehensive command templates and usage examples
- Integrate with existing fullstack and development workflows
Documentation & Safety:
- Update CLAUDE.md with comprehensive Gemini CLI usage guide
- Provide command templates for common analysis patterns
- Include safety measures, path validation, and error handling
- Document integration patterns with existing BMAD workflow
The system enables analysis of large codebases that exceed normal context limits
while maintaining BMAD's safety-first approach and task-based architecture.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add research-coordinator agent with multi-perspective coordination capabilities
- Add adaptable researcher agent with domain-specific specialization profiles
- Implement unified request-research task for agent-to-research communication
- Add research logging and indexing system in docs/research/
- Include web search capabilities and configurable team size (default 3)
- Update analyst, architect, and pm agents with research command integration
- Add research team configuration and integrate into fullstack team
- Implement research synthesis templates and quality assurance checklists
- Add comprehensive research methodologies and domain expertise documentation
The system enables any agent to request specialized research through standardized
task interface, with research coordinator spawning domain experts and synthesizing
findings into actionable reports while maintaining indexed research log.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
* feat: add OpenCode integration implementation plan for BMAD-METHOD
* installer(opencode): add OpenCode target metadata in install.config.yaml
* chore(deps): add comment-json for JSONC parsing in OpenCode integration
* feat(installer/opencode): implement setupOpenCode with minimal instructions merge and BMAD-managed agents/commands
* feat(installer): add OpenCode (SST) to IDE selector and CLI --ide help
* fix(opencode): align generated opencode.json(c) with schema (instructions as strings; agent.prompt; command.template; remove unsupported fields)
* feat(installer): enhance OpenCode setup with agent selection and prefix options
* fix: update configuration file references from `bmad-core/core-config.yaml` to `.bmad-core/core-config.yaml` across multiple agent and task files for consistency and clarity.
* refactor: streamline OpenCode configuration prompts and normalize instruction paths for agents and tasks
* feat: add tools property to agent definitions for enhanced functionality. Otherwise opencode consders the subagents as readonly
* feat: add extraction of 'whenToUse' from agents markdown files for improved agent configuration in opencode
* feat: enhance task purpose extraction from markdown files with improved parsing and cleanup logic
* feat: add collision warnings for non-BMAD-managed agent and command keys during setup
* feat: generate and update AGENTS.md for OpenCode integration with agent and task details
* feat: add compact AGENTS.md generator and JSON-only integration for OpenCode
* chore(docs): remove completed OpenCode integration implementation plans
* feat: enable default prefixes for agent and command keys to avoid collisions
* fix: remove unnecessary line breaks in 'whenToUse' descriptions for QA agents to mathc the rest of the agents definitions and improve programatic parsing of whenToUse prop
* fix: update OpenCode references to remove 'SST' for consistency across documentation and configuration
* fix: update agent mode from 'subagent' to 'all' for consistency in agent definitions
* fix: consolidate 'whenToUse' description format for clarity and consistent parsing
- Change 'Dependencies map to root/type/name' to 'Dependencies map to {root}/{type}/{name}'
- Fixes path resolution issue where 'root' was treated as literal directory name
- Ensures proper variable interpolation for dependency file loading
- Resolves agent inability to find core-config.yaml and other project files
- Alphabetized all commands in agent files while maintaining help first and exit last
- Alphabetized all dependency categories (checklists, data, tasks, templates, utils, workflows)
- Alphabetized items within each dependency category across all 10 core agents:
- analyst.md: commands and dependencies reorganized
- architect.md: commands and dependencies reorganized
- bmad-master.md: commands and dependencies reorganized, fixed YAML parsing issue
- bmad-orchestrator.md: commands and dependencies reorganized
- dev.md: commands and dependencies reorganized
- pm.md: commands and dependencies reorganized
- po.md: commands and dependencies reorganized
- qa.md: commands and dependencies reorganized
- sm.md: commands and dependencies reorganized
- ux-expert.md: commands and dependencies reorganized
- Fixed YAML parsing error in bmad-master.md by properly quoting activation instructions
- Rebuilt all agent bundles and team bundles successfully
- Updated expansion pack bundles including new creative writing agents
This improves consistency and makes it easier to locate specific commands and dependencies
across all agent configurations.
* This PR introduces a powerful new Codebase Flattener Tool that aggregates entire codebases into AI-optimized XML format, making it easy to share project context with AI assistants for analysis, debugging, and development assistance.
- AI-Optimized XML Output : Generates clean, structured XML specifically designed for AI model consumption
- Smart File Discovery : Recursive file scanning with intelligent filtering using glob patterns
- Binary File Detection : Automatically identifies and excludes binary files, focusing on source code
- Progress Tracking : Real-time progress indicators with comprehensive completion statistics
- Flexible Output : Customizable output file location and naming via CLI arguments
- Gitignore Integration : Automatically respects .gitignore patterns to exclude unnecessary files
- CDATA Handling : Proper XML CDATA sections with escape sequence handling for ]]> patterns
- Content Indentation : Beautiful XML formatting with properly indented file content (4-space indentation)
- Error Handling : Robust error handling with detailed logging for problematic files
- Hierarchical Formatting : Clean XML structure with proper indentation and formatting
- File Content Preservation : Maintains original file formatting within indented CDATA sections
- Exclusion Logic : Prevents self-inclusion of output files ( flattened-codebase.xml , repomix-output.xml )
- tools/flattener/main.js - Complete flattener implementation with CLI interface
- package.json - Added new dependencies (glob, minimatch, fs-extra, commander, ora, chalk)
- package-lock.json - Updated dependency tree
- .gitignore - Added exclusions for flattener outputs
- README.md - Comprehensive documentation with usage examples
- docs/bmad-workflow-guide.md - Integration guidance
- tools/cli.js - CLI integration
- .vscode/settings.json - SonarLint configuration
```
current directory
npm run flatten
npm run flatten -- --output my-project.xml
npm run flatten -- -o /path/to/output/codebase.xml
```
The tool provides comprehensive completion summaries including:
- File count and breakdown (text/binary/errors)
- Source code size and generated XML size
- Total lines of code and estimated token count
- Processing progress and performance metrics
- Bug Fix : Corrected typo in exclusion patterns ( repromix-output.xml → repomix-output.xml )
- Performance : Efficient file processing with streaming and progress indicators
- Reliability : Comprehensive error handling and validation
- Maintainability : Clean, well-documented code with modular functions
- AI Integration : Perfect for sharing codebase context with AI assistants
- Code Reviews : Streamlined code review process with complete project context
- Documentation : Enhanced project documentation and analysis capabilities
- Development Workflow : Improved development assistance and debugging support
This tool significantly enhances the BMad-Method framework's AI integration capabilities, providing developers with a seamless way to share complete project context for enhanced AI-assisted development workflows.
* docs(bmad-core): update documentation for enhanced workflow and user guide
- Fix typos and improve clarity in user guide
- Add new enhanced development workflow documentation
- Update brownfield workflow with flattened codebase instructions
- Improve consistency in documentation formatting
* chore: remove unused files and configurations
- Delete deprecated bmad workflow guide and roomodes file
- Remove sonarlint project configuration
- Downgrade ora dependency version
- Remove jest test script
* Update package.json
Removed jest as it is not needed.
* Update working-in-the-brownfield.md
added documentation for sharding docs
* perf(flattener): improve memory efficiency by streaming xml output
- Replace in-memory XML generation with streaming approach
- Add comprehensive common ignore patterns list
- Update statistics calculation to use file size instead of content length
* fix/chore: Update console.log for user-guide.md install path. Cleaned up config files/folders and updated .gitignore (#347)
* fix: Update console.log for user-guide.md install path
Changed
IMPORTANT: Please read the user guide installed at docs/user-guilde.md
to
IMPORTANT: Please read the user guide installed at .bmad-core/user-guide.md
WHY: the actual install location of the user-guide.md is in the .bmad-core directory.
* chore: remove formatting configs and clean up gitignore
- Delete husky pre-commit hook and prettier config files
- Remove VS Code chat/copilot settings
- Reorganize and clean up gitignore entries
* feat: Overhaul and Enhance 2D Unity Game Dev Expansion Pack (#350)
* Updated game-sm agent to match the new core framework patterns
* feat:Created more comprehensive game story matching new format system as well
* feat:Added Game specific course correct task
* feat:Updated dod-checklist to match new DoD format
* feat:Added new Architect agent for appropriate architecture doc creation and design
* feat:Overhaul of game-architecture-tmpl template
* feat:Updated rest of templates besides level which doesnt really need it
* feat: Finished extended architecture documentation needed for new game story tasks
* feat: Updated game Developer to new format
* feat: Updated last agent to new format and updated bmad-kb. bmad-kb I did my best with but im not sure of it's valid usage in the expansion pack, the AI generated more of the file then myself. I made sure to include it due to the new core-config file
* feat: Finished updating designer agent to new format and cleaned up template linting errors
* Built dist for web bundle
* Increased expansion pack minor verison number
* Updated architecht and design for sharding built-in
* chore: bump bmad-2d-unity-game-dev version (minor)
* updated config.yaml for game-specific pieces to supplement core-config.yaml
* Updated game-core-config and epic processing for game story and game design. Initial implementation was far too generic
* chore: bump bmad-2d-unity-game-dev version (patch)
* feat: Fixed issue with multi-configs being needed. chore: bump bmad-2d-unity-game-dev version (patch)
* Chore: Built web-bundle
* feat: Added the ability to specify the unity editor install location.\nchore: bump bmad-2d-unity-game-dev version (patch)
* feat: core-config must be in two places to support inherited tasks at this time so added instructions to copy and create one in expansion pack folder as well. chore: bump bmad-2d-unity-game-dev version (patch)
* This PR introduces a powerful new Codebase Flattener Tool that aggregates entire codebases into AI-optimized XML format, making it easy to share project context with AI assistants for analysis, debugging, and development assistance.
- AI-Optimized XML Output : Generates clean, structured XML specifically designed for AI model consumption
- Smart File Discovery : Recursive file scanning with intelligent filtering using glob patterns
- Binary File Detection : Automatically identifies and excludes binary files, focusing on source code
- Progress Tracking : Real-time progress indicators with comprehensive completion statistics
- Flexible Output : Customizable output file location and naming via CLI arguments
- Gitignore Integration : Automatically respects .gitignore patterns to exclude unnecessary files
- CDATA Handling : Proper XML CDATA sections with escape sequence handling for ]]> patterns
- Content Indentation : Beautiful XML formatting with properly indented file content (4-space indentation)
- Error Handling : Robust error handling with detailed logging for problematic files
- Hierarchical Formatting : Clean XML structure with proper indentation and formatting
- File Content Preservation : Maintains original file formatting within indented CDATA sections
- Exclusion Logic : Prevents self-inclusion of output files ( flattened-codebase.xml , repomix-output.xml )
- tools/flattener/main.js - Complete flattener implementation with CLI interface
- package.json - Added new dependencies (glob, minimatch, fs-extra, commander, ora, chalk)
- package-lock.json - Updated dependency tree
- .gitignore - Added exclusions for flattener outputs
- README.md - Comprehensive documentation with usage examples
- docs/bmad-workflow-guide.md - Integration guidance
- tools/cli.js - CLI integration
- .vscode/settings.json - SonarLint configuration
```
current directory
npm run flatten
npm run flatten -- --output my-project.xml
npm run flatten -- -o /path/to/output/codebase.xml
```
The tool provides comprehensive completion summaries including:
- File count and breakdown (text/binary/errors)
- Source code size and generated XML size
- Total lines of code and estimated token count
- Processing progress and performance metrics
- Bug Fix : Corrected typo in exclusion patterns ( repromix-output.xml → repomix-output.xml )
- Performance : Efficient file processing with streaming and progress indicators
- Reliability : Comprehensive error handling and validation
- Maintainability : Clean, well-documented code with modular functions
- AI Integration : Perfect for sharing codebase context with AI assistants
- Code Reviews : Streamlined code review process with complete project context
- Documentation : Enhanced project documentation and analysis capabilities
- Development Workflow : Improved development assistance and debugging support
This tool significantly enhances the BMad-Method framework's AI integration capabilities, providing developers with a seamless way to share complete project context for enhanced AI-assisted development workflows.
* chore: remove unused files and configurations
- Delete deprecated bmad workflow guide and roomodes file
- Remove sonarlint project configuration
- Downgrade ora dependency version
- Remove jest test script
* docs: update command names and agent references in documentation
- Change `*create` to `*draft` in workflow guide
- Update PM agent commands to use consistent naming
- Replace `analyst` references with `architect`
- Fix command examples to match new naming conventions
---------
Co-authored-by: PinkyD <paulbeanjr@gmail.com>
This release introduces significant enhancements across multiple areas:
QA Agent Transformation:
- Transform QA agent into senior developer role with active code refactoring abilities
- Add review-story task enabling QA to review, refactor, and improve code directly
- Integrate QA review step into standard development workflow (SM → Dev → QA)
- QA can fix small issues directly and leave checklist for remaining items
- Updated dev agent to maintain File List for QA review focus
Knowledge Base Improvements:
- Add extensive brownfield development documentation and best practices
- Clarify Web UI vs IDE usage with cost optimization strategies
- Document PRD-first approach for large codebases/monorepos
- Add comprehensive expansion packs explanation
- Update IDE workflow to include QA review step
- Clarify agent usage (bmad-master vs specialized agents)
Brownfield Enhancements:
- Create comprehensive Working in the Brownfield guide
- Add document-project task to analyst agent capabilities
- Implement PRD-first workflow option for focused documentation
- Transform document-project to create practical brownfield architecture docs
- Document technical debt, workarounds, and real-world constraints
- Reference actual files instead of duplicating content
- Add impact analysis when PRD is provided
Documentation Task Improvements:
- Simplify to always create ONE unified architecture document
- Add deep codebase analysis phase with targeted questions
- Focus on documenting reality including technical debt
- Include Quick Reference section with key file paths
- Add practical sections: useful commands, debugging tips, known issues
Workflow Updates:
- Update all 6 workflow files with detailed IDE transition instructions
- Add clear SM → Dev → QA → Dev cycle explanation
- Emphasize Gemini Web for brownfield analysis (1M+ context advantage)
- Support both PRD-first and document-first approaches
This release significantly improves the brownfield development experience and introduces a powerful shift-left QA approach with senior developer mentoring.
Issue Being Solved:
Dev agent was marking tasks as complete even when tests/lint/typecheck failed,
causing broken code to reach merge and creating technical debt.
Gap in System:
Missing explicit quality gates in dev agent configuration to block task completion
until all automated validations pass.
Solution:
- Add "Quality Gate Discipline" core principle
- Update task execution flow: Execute validations→Only if ALL pass→Update [x]
- Add "Failing validations" to blocking conditions
- Simplify completion criteria to focus on validation success
Alignment with BMAD Core Principles:
- Quality-First: Prevents defective code progression through workflow
- Agent Excellence: Ensures dev agent maintains high standards
- Technology Agnostic: Uses generic "validations" concept vs specific tools
Small, focused change that strengthens existing dev agent without architectural changes.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-authored-by: Claude <noreply@anthropic.com>
- Remove 'root' property from YAML when building web bundles
- Remove 'IDE-FILE-RESOLUTION' and 'REQUEST-RESOLUTION' properties
- Filter out IDE-specific activation instructions
- Keep agent header minimal for web bundles
- Ensures web bundles are clean of IDE-specific configuration
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added two concise activation instructions to SM agent
- IDE-FILE-RESOLUTION: One-line explanation of file path mapping
- REQUEST-RESOLUTION: One-line instruction for flexible request matching
- Simplified file-resolution-context.md to be a quick reference
- Removed verbose documentation in favor of clear, actionable instructions
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added file resolution section to SM agent explaining path patterns
- Created reusable file-resolution-context.md utility
- Documents how agents resolve tasks/templates/checklists to file paths
- Provides natural language to command mapping examples
- Helps IDE agents understand file system structure
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
LNKB82REZ2GE4
Javier Gomez
Daniel Dabrowski
Piatra Automation
Wang-tianhao
Brian Madison
manjaroblack
Murat K Ozcan
Duane Cilliers
gabadi