- Updated intelligent-tools-strategy.md to include `--skip-git-repo-check` for Codex write access and development commands.
- Improved context gathering and analysis processes in mcp-tool-strategy.md with additional examples and guidelines for file searching.
- Introduced new command concept-enhanced.md for enhanced intelligent analysis with parallel CLI execution and design blueprint generation.
- Added context-gather.md command for intelligent collection of project context based on task descriptions, generating standardized JSON context packages.
- Implemented `/workflow:plan` for creating detailed implementation plans with task decomposition and context gathering.
- Added `/workflow:resume` for intelligent session resumption with automatic progress analysis.
- Introduced `/workflow:review` for executing the final phase of quality validation and generating review reports.
- Developed `/workflow:status` to provide on-demand views of workflow status and task progress.
- Created `/workflow:test-gen` to generate comprehensive test workflows based on completed implementation tasks, ensuring full test coverage.
## New Features
- **MCP Tools Integration**: Added support for Model Context Protocol tools
- Exa MCP Server: External API patterns and best practices
- Code Index MCP: Advanced internal codebase search and indexing
- **Enhanced Workflow Planning**: Updated pre_analysis to include MCP tool steps
- **Documentation Updates**: Added MCP tool setup guides and usage examples
## Changes
### Core Components
- Updated `plan.md` with MCP integration principles and implementation approach guidelines
- Added MCP tool steps in pre_analysis workflow: `mcp_codebase_exploration`, `mcp_external_context`
- Enhanced context accumulation with external best practices lookup
### Documentation
- Added comprehensive MCP tools section in both English and Chinese README
- Updated installation requirements and integration guidelines
- Added GitHub repository links for required MCP servers
### Agent Enhancements
- Updated multiple agents to support MCP tool integration
- Enhanced context gathering capabilities with external pattern analysis
## Technical Details
- MCP tools provide faster analysis through direct codebase indexing
- Automatic fallback to traditional bash/CLI tools when MCP unavailable
- Enhanced pattern recognition and similarity detection capabilities
🧪 **Experimental**: MCP integration is currently experimental and optional
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Update agent definitions with enhanced context and flow control
- Standardize command templates for consistent CLI tool integration
- Improve gemini-wrapper with better token management and path handling
- Simplify qwen-wrapper for streamlined execution
- Enhance intelligent-tools-strategy with directory navigation patterns
- Add proper quoting for Windows path compatibility in all commands
- Update workflow planning documentation with current architectural insights
These changes align the command system with the four-layer architecture
analysis and improve cross-platform compatibility.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Changes:
- Removed pycli backend references from README.md
- Removed pycli backend references from README_CN.md
- Deleted obsolete python-tools-strategy.md file
- Updated project structure documentation to reflect current scripts
- Replaced Python backend section with enhanced test generation features
This cleanup ensures documentation accurately reflects the current system
without the removed Python CLI components.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Major updates:
- Enhanced workflow lifecycle with 6-phase development process
- Updated README.md with comprehensive CLI command documentation
- Updated README_CN.md with Chinese documentation reflecting v2.0 features
- Added Qwen CLI commands and improved Gemini/Codex command descriptions
- Enhanced brainstorming role commands and workflow session management
- Updated integration requirements for all three CLI tools
Script cleanup:
- Removed unused Python CLI scripts (install_pycli.sh, pycli, pycli.conf)
- Removed deprecated path reading scripts (read-paths.sh, read-task-paths.sh)
- Removed tech-stack-loader.sh
- Kept core scripts: gemini-wrapper, qwen-wrapper, get_modules_by_depth.sh
Architecture improvements:
- JSON-First data model as single source of truth
- Atomic session management with marker files
- Multi-agent coordination for complex task execution
- Role-based brainstorming with synthesis capabilities
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add concept-eval.md command for concept evaluation workflow
- Add concept-eval.txt template for structured concept analysis
- Enhance plan.md with required IMPL_PLAN.md structure format
- Define standardized format for project planning documentation
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add Qwen as third CLI tool for architecture analysis and code generation
- Update decision framework to include architecture/code generation workflow
- Enhance tool selection matrix with Qwen specialization areas
- Add Qwen command templates and directory context support
- Update workflow integration with dedicated architecture phase
- Provide comprehensive examples for all three tools (Gemini/Qwen/Codex)
- Clarify tool responsibilities and use cases
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add -C parameter to gemini-wrapper for directory change before analysis
- Update intelligent-tools-strategy.md to use -C (uppercase) consistently
- Replace cd [directory] functionality with -C parameter throughout docs
- Add directory context examples for both Gemini and Codex tools
- Enhance context optimization strategy with -C parameter usage
- Test confirms -C parameter works correctly with token analysis in target directory
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Added context optimization to decision principles
- Include guidance for using cd [directory] to reduce irrelevant context
- Added examples showing focused vs multi-scope analysis patterns
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Create /workflow:verify command for plan validation before execution
- Implement gemini strategic analysis and codex technical analysis
- Add cross-validation framework with user approval workflow
- Enhance brainstorm auto.md with comprehensive pre_analysis configuration
- Include specialized verification prompt templates for structured analysis
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Transform from command reference to strategic decision framework
- Add core framework with decision principles and execution rules
- Include universal command template with PURPOSE/TASK/CONTEXT/EXPECTED structure
- Add tool selection matrix and workflow integration patterns
- Include planning checklist and best practices sections
- Add hierarchical vector system details and migration benefits
- Provide quick setup guide and decision framework
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Modify install_pycli.sh to add ~/.claude/scripts/ to PATH instead of creating alias
- Remove redundant alias configuration, use direct PATH registration
- Update documentation to reflect simplified installation process
- Delete README.md as requested by user
- Now users can call 'pycli' directly after installation without alias setup
Installation process:
1. Run install_pycli.sh
2. Reload shell configuration (source ~/.bashrc)
3. Use 'pycli' command directly
🤖 Generated with Claude Code
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace markdown template with Claude Code TodoWrite tool usage
- Update documentation to use built-in TodoWrite API instead of manual TODO_LIST.md updates
- Align with JSON-only data model and real-time progress tracking principles
- Add proper TodoWrite integration rules and examples
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Convert direct agent name references to @agent-name format in documentation
- Preserve Task() function calls and subagent_type parameters unchanged
- Keep task content agent identifiers in original format (for display only)
- Maintain document reference paths in original format
- Update workflow architecture, task management, and execution files
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Clarify that 20-minute timeout applies specifically to bash() wrapped commands
- Add explicit examples of bash command format with gemini-wrapper and codex
- Align timeout description with actual usage in workflow-architecture.md
- Remove ambiguity about which CLI commands use the default timeout
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Remove -t 1200000 timeout parameters from gemini-wrapper and codex commands
- Use default timeout behavior as defined in intelligent-tools-strategy.md
- Simplify command templates for better readability
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add pre-analysis documentation check as first step
- Reference specific paths from /workflow:docs output structure
- Update flow_control to selectively load relevant documentation
- Remove outdated plan-deep command
- Add new doc-generator agent and workflow files
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Create context-search-strategy.md with comprehensive search command reference
- Include rg, find, and grep usage patterns for efficient codebase exploration
- Add workflow-specific search examples for task analysis and session management
- Reference new strategy in CLAUDE.md CLI tool protocols section
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Enable multiple active workflow sessions to coexist
- Add session detection logic for multiple .active-* markers
- Update execute command to prompt user selection when multiple sessions found
- Update plan command with multi-session awareness
- Maintain session isolation while allowing parallel workflow instances
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add core execution rules with 20-minute default timeout for CLI commands
- Introduce CONTEXT field to templates for file references and session memory
- Update all command examples with CONTEXT field usage
- Enhance planning checklist with context gathering step
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Reorganize content into logical sections: Core Framework, Tool Specifications, Decision Matrix, Workflow Integration, Usage Patterns
- Add structured prompt templates with PURPOSE|TASK|EXPECTED format for both Gemini and Codex
- Consolidate redundant sections while preserving all unique content
- Reduce file size by 13% (221→193 lines) through better organization
- Enhance workflow integration rules with mandatory planning process
- Improve decision matrix with task categorization and frequency guidelines
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Updated intelligent-tools-strategy.md to include -s parameter in all codex examples
- Added comprehensive sandbox modes documentation
- Updated tools-implementation-guide.md with -s workspace-write throughout
- Clarified that -s workspace-write is recommended for development tasks
- Added sandbox mode explanations: read-only, workspace-write, danger-full-access
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- plan.md: Compress from 307 to 183 lines while preserving all core functionality
- Add project structure analysis with get_modules_by_depth.sh integration
- Implement 5-field JSON schema compliance with workflow-architecture.md
- Add detailed context acquisition strategy with tool templates
- Include comprehensive file cohesion principles and variable system
- Maintain 10-task limit enforcement and pre-planning analysis requirements
- execute.md: Compress from 380 to 220 lines with enhanced agent context
- Preserve complete Task execution patterns with full flow control context
- Maintain comprehensive session context, implementation guidance, and error handling
- Streamline discovery process while keeping all critical execution details
- Keep complete agent assignment and status management functionality
- workflow-architecture.md: Minor structural updates for consistency
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Simplified argument hints and examples in plan.md
- Enhanced input detection logic for files, issues, and text
- Introduced analysis levels for improved context analysis
- Updated core rules for task limits and decomposition strategies
- Improved session management with active session marker system
- Expanded file structure reference and naming conventions
- Clarified task hierarchy and status rules
- Added detailed validation and error handling procedures
- Streamlined document templates for implementation plans and task summaries
- Add critical warnings in codex-unified.md that no wrapper script exists
- Clarify in intelligent-tools.md that Gemini has wrapper, Codex uses direct commands
- Prevent confusion about non-existent ~/.claude/scripts/codex
- Emphasize correct usage: gemini-wrapper vs codex --full-auto exec
- Clean up CLAUDE.md tool references for consistency
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Create intelligent-tools.md as strategic guide for tool selection
- Reference intelligent-tools.md from CLAUDE.md for global access
- Add three analysis levels to plan.md (quick/standard/deep)
- Separate tool selection strategy from plan command implementation
- Maintain clear separation of concerns between strategy and execution
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Add detailed mermaid workflow diagrams in WORKFLOW_DIAGRAMS.md
- Update README.md and README_CN.md with v1.2 features and architecture visualization
- Enhance system architecture diagrams with CLI routing and agent coordination flows
- Document major enhancements since v1.0: task saturation control, Gemini wrapper intelligence
- Add command execution flow diagrams and comprehensive workflow visualizations
- Update CLI guidelines in codex-unified.md and gemini-unified.md with bash() syntax
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
Updated all CLI command examples across agent documentation and workflow
guides to use bash() wrapper for proper tool execution:
- Modified action-planning-agent.md CLI usage standards
- Updated code-developer.md analysis CLI commands
- Enhanced conceptual-planning-agent.md execution logic
- Revised code-review-test-agent.md CLI commands
- Wrapped all gemini-wrapper calls in gemini-unified.md
- Updated all codex commands in codex-unified.md
This ensures consistent tool execution patterns across all documentation
and provides clear guidance for proper CLI tool invocation.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
CONSISTENCY IMPROVEMENTS:
- Update gemini-unified.md examples to use full path ~/.claude/scripts/gemini-wrapper
- Add setup note indicating script auto-installs to ~/.claude/scripts/ location
- Clarify usage instructions to emphasize full path usage
AGENT DOCUMENTATION UPDATES:
- action-planning-agent.md: Use full path for gemini-wrapper
- code-developer.md: Use full path for gemini-wrapper
- code-review-test-agent.md: Use full path for gemini-wrapper
- conceptual-planning-agent.md: Use full path for gemini-wrapper
BENEFITS:
- Prevents "command not found" errors
- Ensures consistent invocation across all documentation
- Makes path explicit and unambiguous for users
- Supports direct copy-paste usage from documentation
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
ENHANCEMENTS:
- Create gemini-wrapper script with automatic token counting and smart flag management
- Auto-add --approval-mode based on task type (analysis=default, execution=yolo)
- Raise token threshold to 2M for better large project handling
- Add comprehensive parameter documentation for --approval-mode and --include-directories
WRAPPER FEATURES:
- Token-based --all-files management (small projects get --all-files automatically)
- Smart task detection for approval modes
- Error logging to ~/.claude/.logs/gemini-errors.log
- Complete parameter passthrough for full gemini compatibility
DOCUMENTATION UPDATES:
- Update gemini-unified.md with wrapper usage guidelines and examples
- Add intelligent wrapper as recommended approach
- Document all agent files to use gemini-wrapper instead of direct gemini calls
- Include new parameter reference and best practices
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Update unified file structure to stress on-demand creation principle
- Add creation strategy section with explicit guidance:
- Initial setup creates only required files (session.json, IMPL_PLAN.md, TODO_LIST.md, .task/)
- Optional directories (.brainstorming/, .chat/, .summaries/) created when first needed
- Update core design principles to include on-demand file creation
- Add session initialization and directory creation examples in Data Operations
This ensures workflows start lean and expand only as functionality is used.
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
- Replace separate Main Tasks/Subtasks sections with unified hierarchical list
- Use ▸ symbol for container tasks (with subtasks) instead of checkboxes
- Maintain standard - [ ]/- [x] for executable leaf tasks
- Add 2-space indentation to show task hierarchy clearly
- Include status legend for better user understanding
Benefits:
- Eliminates confusion about non-executable main tasks
- Provides clear visual hierarchy in single list
- Reduces TODO_LIST complexity and improves usability
- Aligns container task concept with execution model
🤖 Generated with [Claude Code](https://claude.ai/code)
Co-Authored-By: Claude <noreply@anthropic.com>
