Features
Self-evolving system that learns from sessions
Event-driven hooks for lifecycle moments
Convention lock via CLAUDE.md rules
Automatic skill injection into prompts
Smart permission gate using LLM evaluation
Context delegation to sub-agents
Persistent memory with staleness detection
Daily session logs for full timeline
Bidirectional task sync with Claude's task system
Complexity-based routing for task sizing
Automated quality checks (TypeScript, ESLint, builds)
Daily security scan for vulnerabilities
Performance audit with speed scores and loading times
Dependency audit every 3 days
Error triage with fix recommendations
Multi-agent orchestration (multiple AI agents)
File editing via VS Code extension
Terminal control and command execution
Git integration for version-aware assistance
Autonomous mode (coming soon)
Advanced task routing rules
Local processing for privacy
Customizable model deployment
