
Skill Inspector
Reviews other AI agent skills for safety before you install them, using a static scanner plus manual source review.
- Repository
- NVIDIA/SkillSpector
- Stars
- 23,933
- Licence
- Apache-2.0
- Files
- 1 file
What it does
Once installed, this skill tells the agent to act as a security reviewer for other agent skills. Before you install or keep a skill, the agent runs NVIDIA's SkillSpector scanner on it (without using an LLM by default), reads the source code and instructions, checks whether the behavior matches the stated purpose, and returns a verdict of APPROVE, CAUTION, or REJECT with the evidence behind it.
Great for
- Developers who download third-party agent skills and want a safety check first
- Teams that review skills before adding them to a shared internal library
- Anyone asking whether a downloaded skill folder looks malicious or over-permissioned
Skip it if
- Guaranteeing a skill is safe: it is a triage aid, and a static scan plus AI judgment can miss risks (the skill itself says a low score can miss semantic risk)
- Reviewing skills that need the scanner but where SkillSpector is not installed; the skill says to continue with manual review only, which is slower and less reliable
Try asking
- “Is this downloaded skill folder safe to install before I add it to Claude?”
- “Review the skill at this GitHub URL for over-permissioned tools or hidden network calls.”
- “Check whether this skill's scripts do anything beyond what its description promises.”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the Apache-2.0 licence and an attribution note.
skill-inspector.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip skill-inspector.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip skill-inspector.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Give the agent a local path or repository URL; for URLs it clones into a temporary directory before review.
- Ask for the full report if you want the rule IDs and file and line evidence, not just the verdict.
- Treat CAUTION as a prompt to read the flagged files yourself before installing.
What's inside
- SKILL.md
Redistributed unmodified from NVIDIA/SkillSpector under the Apache-2.0 licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.