skill-engineer
chunhualiao 제작
Coding Agents & IDEs요약
Design, test, review, and maintain agent skills for OpenClaw systems using multi-agent iterative refinement.
설명
이 스킬이 하는 일
Manages the full lifecycle of agent skills in OpenClaw systems. Three independent subagents handle design, review, and testing separately to prevent a designer from evaluating their own work. Supports iterative refinement loops with configurable quality gates before a skill is considered ready to ship.
Enforcing separate subagents for design and evaluation catches quality gaps that self-review reliably misses, a problem the skill's architecture was built to address after a real 70-minute pipeline stall incident.
사용 시점
- Designing a new skill for an OpenClaw agent kit from scratch
- Running a quality audit across all skills in an agent kit
- Testing whether a skill triggers correctly on sample prompts
- Refactoring an existing skill after a process change or model update
- Evaluating whether a capability-uplift skill has become redundant
예시 워크플로
User asks: Design a skill for analyzing competitor websites
- 1
Gather requirements: target agent, inputs (competitor URLs, analysis criteria), outputs (comparison table, markdown report), and constraints
- 2
Query DeepWiki to verify current OpenClaw API behavior before writing any skill content
- 3
Spawn Designer subagent to generate SKILL.md, skill.yml, README.md, and tests/
- 4
Spawn Reviewer subagent to score the artifacts against the 33-check quality rubric
- 5
Spawn Tester subagent to run self-play validation and measure trigger accuracy
Validated skill directory with all required artifacts and a quality scorecard in README.md, git-committed and ready for publishing
요구 사항
- deepwiki skill installed from liaosvcaf/openclaw-skill-deepwiki
- clawhub CLI installed via npm (npm i -g clawhub) for publishing to ClawhHub registry
설치
clawhub install chunhualiao/skill-engineerCLI 명령을 실행하거나, 프롬프트를 Claude Code, Cursor, Codex 등 사용하는 코딩 에이전트에 붙여넣으세요.