review-orchestrator
由 leegitw 提供
Coding Agents & IDEs简介
Get multiple perspectives on your work — coordinate reviews across cognitive modes.
描述
该技能的功能
Coordinates multiple review perspectives on code, architecture, or documentation within a single skill. Runs technical and creative angles in parallel (twin review), applies cognitive modes like analyzer, architect, and implementer, and enforces quality gates on staged work. No external APIs are called; all review logic runs through the agent's configured model.
Running the same code through distinct analytical perspectives (conflict detection, structural redesign, implementation path) in one command catches more issues than a single-pass review without requiring separate manual review sessions.
何时使用
- Reviewing an auth handler refactor from both technical and UX angles
- Catching API design inconsistencies before publishing a spec
- Blocking a release stage when test coverage drops too far
- Analyzing architecture docs for structural conflicts before a refactor
- Getting concrete implementation steps alongside trade-off analysis on a feature
示例工作流
User asks: /ro twin src/handlers/auth.go
- 1
Parse the target file path and determine which perspectives to apply (technical, creative)
- 2
Run technical perspective review focusing on architecture, patterns, and security
- 3
Run creative perspective review focusing on UX and error message clarity
- 4
Aggregate findings from both perspectives, tagged by severity
- 5
Write results to docs/reviews/ and return a verdict with any required conditions
Multi-perspective review report listing severity-tagged findings from both perspectives, with a verdict of Approved, Approved with conditions, or Blocked
安装
clawhub install leegitw/review-orchestrator运行 CLI 命令,或把这段提示词粘贴到 Claude Code、Cursor、Codex 或你在用的任意编程智能体里。