self-review
由 leic8959-sudo 提供
Communication简介
Automatically review agent output quality before sending to user.
描述
该技能的功能
Runs a rule-based quality check on agent output before it reaches the user. Evaluates clarity, conciseness, actionability, and structure. Returns exit code 0 for approved output or 1 when improvement is needed.
Catches output quality issues without LLM API calls, keeping review costs at zero.
何时使用
- Catching verbose or unclear agent responses before delivery
- Filtering low-quality output in automated agent pipelines
- Running post-generation checks without extra API costs
- Enforcing response structure standards across agent workflows
- Validating output during CI/CD for agent-based tools
示例工作流
User asks: Check this agent response before sending it
- 1
Pipe the generated response text to self-review/index.js via stdin
- 2
Rule-based engine evaluates clarity, conciseness, actionability, and structure
- 3
Returns exit code 0 if output passes all checks
- 4
Returns exit code 1 with feedback if output needs improvement
Exit code 0 with approved response, or exit code 1 indicating which checks failed
安装
clawhub install leic8959-sudo/self-review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。