AI agents write code. Ritsu checks it.
When Claude Code or Cursor generates code, Ritsu catches what the agent misses — scope creep, security vulnerabilities, broken contracts, architecture drift, credential leaks. With session recovery, multi-agent orchestration, and visual UI checks.
AI 写代码，Ritsu 管质量。
Quick Start • Why • Install
# Try it now — no project needed
npx skills add 3kaiu/Ritsu -a claude-code -g -y
ritsu bootstrap --demo
ritsu violations --open| Command | What it does |
|---|---|
/r-think |
Analyze + design contracts |
/r-dev |
Code with policy enforcement |
/r-review |
Quality + security review |
/r-deploy |
Deploy plan + rollback |
/r-hunt |
Root cause diagnosis |
/r-augment |
Test gap analysis |
AI coding agents have three blind spots:
- No quality awareness — they don't check for security issues, architecture drift, or contract compliance
- No memory — when a session ends, everything resets
- No audit trail — no record of what was checked or what it cost
Ritsu is a governance layer that runs alongside AI agents, continuously checking, recording, and recovering.
What it catches: scope creep, XSS/SQL injection, hardcoded credentials, broken API contracts, circular dependencies, console.log/debugger left in production, version mismatches, unauthorized file changes.
Agent writes code → Policy engine (12 detectors / 21 anti-patterns) → Quality gates → Violation tracked
→ Session checkpoint → Contract verified → Visual check
ritsu doctor # Health / 健康检查
ritsu report # Quality analytics / 质量分析
ritsu report --cost # Cost breakdown / 成本追踪
ritsu violations # Open issues / 未解决违规
ritsu trace <id> # Trace tree / 追踪链路
ritsu violations --trend # Monthly trend / 月度趋势MIT © 2024-2026 3kaiu