Skill Refiner
v3.0.0Quality audit for OpenClaw skill files — scores each SKILL.md on 5 dimensions (clarity, completeness, authenticity, examples, freshness), finds broken refere...
⭐ 0· 258·0 current·0 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Benign
high confidencePurpose & Capability
The name/description match the runtime instructions (it audits SKILL.md files, scores quality, and writes a report). Minor metadata mismatch: the registry lists no required binaries, but SKILL.md explicitly expects grep, curl, and python3 — these are reasonable for the stated task but should be declared in the registry metadata.
Instruction Scope
Instructions stay within scope: they read skills/*/SKILL.md, analyze frontmatter/content/code blocks/references, and produce a report. The only external action is an optional link freshness check (HTTP requests). The skill documents 'read-only' behavior and explicitly forbids editing certain files (SOUL.md, MEMORY.md, AGENTS.md).
Install Mechanism
No install spec and no code files — lowest-risk model. The skill is instruction-only and does not download or execute external archives.
Credentials
The skill requests no credentials or config paths. It operates on repository files and creates a local report; no secrets or unrelated env vars are required.
Persistence & Privilege
always:false (normal). It writes a report to memory/skill-refiner-log.md which is persistent within the agent workspace; users should confirm that storing aggregated review output there is acceptable. Autonomous invocation is allowed by default but not excessive for this utility.
Assessment
This skill appears to do what it says: audit SKILL.md files and save a local markdown report. Before installing, confirm your agent environment has grep, curl, and python3 available (the SKILL.md expects them though metadata doesn't list them). Be aware it will read every skills/*/SKILL.md in the workspace and append reports to memory/skill-refiner-log.md (so review that file for any sensitive content you don't want aggregated). The link-check step is optional but will make outbound HTTP requests — disable it if you want a fully offline review. Run it manually once to verify behavior and check that saving reports to the 'memory' location aligns with your privacy/policy requirements.Like a lobster shell, security has layers — review code before you run it.
latestvk97cw8zbew75kb85a4d5sdk47h837x29
License
MIT-0
Free to use, modify, and redistribute. No attribution required.