Skip to content

Release v1.4.0

Choose a tag to compare

View all tags
@shinpr shinpr released this 03 Aug 03:32
· 182 commits to main since this release
v1.4.0
ab33a51
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

Overview

This release introduces a comprehensive dynamic rule selection system that significantly improves AI agent efficiency and accuracy. The system reduces rule content by 15-18% while maintaining or improving functionality through intelligent rule selection and enhanced agent capabilities.

Major Features

🎯 Dynamic Rule Selection System

  • New rules-index.yaml files in both docs/rules-ja/ and docs/rules-en/
    • Provides metadata for each rule file (description, priority, keywords)
    • Enables AI agents to select only relevant rules based on task context
    • Optimizes context window usage for better performance

🤖 New rule-advisor Subagent

  • Intelligent rule selection agent that maximizes AI execution accuracy
  • Selects minimal effective ruleset based on task requirements
  • Provides structured guidance including:
    • Task essence understanding
    • Applicable rules selection
    • Past failure pattern recognition
    • First action recommendations

📝 Rule Content Optimization

  • Japanese rules: Reduced by 353 lines (17.7%) / 5,330 characters (13.9%)
  • English rules: Reduced by 353 lines (17.7%) / 9,586 characters (15.5%)
  • CLAUDE.ja.md: Reduced by 39 lines (22.3%) / 1,040 characters (25.9%)
  • CLAUDE.en.md: Reduced by 38 lines (21.8%) / 2,655 characters (32.1%)
  • Removed canonical-phrases.md - content integrated into other files

📚 Enhanced Documentation

  • Updated README files with rule index system explanation
  • Added detailed descriptions for all 10 subagents
  • Documented the dynamic rule selection benefits
  • Improved project structure documentation

Technical Improvements

Rule System Enhancements

  • Mandatory integration between TodoWrite and rule-advisor
  • Automatic rule selection before task execution
  • Meta-cognition support through structured task analysis
  • Past failure pattern recognition and avoidance

Agent Behavior Controls

  • Stricter implementation approval requirements
  • Automatic stops for large-scale changes (5+ files)
  • Forced rule-advisor execution on repeated errors
  • Enhanced progress tracking through TodoWrite integration

Breaking Changes

  • canonical-phrases.md has been removed - content integrated into other rule files
  • Rule loading now prioritizes the rule index system over loading all files

Migration Guide

  1. Update language settings: npm run lang:en or npm run lang:ja
  2. Review new rules-index.yaml files for rule organization
  3. Utilize rule-advisor for optimal rule selection in tasks
  4. Update any custom scripts that reference canonical-phrases.md

Commit History

  • 6872f86 feat: Release v1.4.0 - Enhanced documentation and rule-advisor integration
  • ec766f5 feat: Complete English translation of rule files from Japanese sources
  • 29a418e feat: Complete English translation of agent definitions and commands
  • 0dc7b63 refactor(rules): Rule system evaluation and optimization
  • eac63ea refactor(rules): CLAUDE.ja.md accuracy improvement and rule-advisor integration
  • 5001d38 refactor: Complete migration to dynamic rule selection system

Statistics

  • Total rule content reduction: ~15-18% across all languages
  • 10 specialized subagents now available
  • 6 core rule files (down from 7)
  • Full bilingual support (Japanese/English)

Future Considerations

  • Further optimization opportunities exist in consolidating common concepts across files
  • Potential for more granular rule selection based on specific task types
  • Enhanced integration between agents for complex workflows

This release represents a significant step forward in making AI-assisted development more efficient and accurate through intelligent rule management and enhanced agent capabilities.

Assets 2
Loading