Skill Reviewer

Review and audit agent skills (SKILL.md files) for quality, correctness, and effectiveness. Use when evaluating a skill before publishing, reviewing someone else's skill, scoring skill quality, identifying defects in skill content, or improving an existing skill.

This instruction-only skill appears to do what it says: it contains a detailed checklist and scoring rubric for auditing SKILL.md files. The main red flag is the metadata requiring the npx binary even though the instructions never call npx or any Node tooling. Before installing or publishing, ask the author to explain why npx is required (for example, if they expect the reviewer to run linters or formatters via npx) and update the SKILL.md to document any external commands the agent will run. If there's no real need for npx, request that the metadata be removed — unnecessary declared binaries broaden the skill's capabilities without justification. As always, run this skill in a constrained or sandboxed environment if you want to be extra cautious.

Type: OpenClaw Skill Name: skill-reviewer Version: 1.0.0 The 'skill-reviewer' skill is designed to audit other OpenClaw skills. Its instructions and suggested commands are entirely aligned with this stated purpose, focusing on structural checks, content quality assessment, and review workflows. The skill uses standard shell commands like `head`, `grep`, and `cat` for file inspection, and `npx molthub@latest install skill-name` for installing skills to be reviewed. While `npx` can execute external code, in this context, it is used transparently to facilitate the core function of skill review, not for malicious execution or data exfiltration by this skill itself. There is no evidence of prompt injection to subvert the agent's intent or perform unauthorized actions.