skillspector
Skill0 starsChanged 2 months agoMITView on GitHub
Security-scan an AI agent skill before installing or shipping it. Use when asked to audit/scan/vet a skill, plugin, or MCP for safety; when installing an untrusted skill from a marketplace/GitHub; or before publishing a skill. Wraps NVIDIA SkillSpector — 68 vulnerability patterns across 17 categories (prompt injection, data exfiltration, privilege escalation, supply chain, tool poisoning, MCP least-privilege, etc.), two-stage static + optional LLM analysis, OSV.dev CVE lookups, SARIF/JSON/Markdown reports.
- Reads credentials
2 other projects use a version of this.
Adopted by
Most recently taken up by
- brenorb/freedom-skillstheir own version, first seen Oct 27, 2026
- drewmerc302/claude-skillstheir own version, first seen Oct 13, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- brenorb/freedom-skills★ 20same name
- drewmerc302/claude-skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository20 other files it gives its agents.
- agent-harness-constructionSkill
- claudex-loopSkill
- click-path-auditSkill
- context-budgetSkill
- councilSkill
- deploy-projectSkill
- dreamSkill
- loop-design-checkSkill
- rebootSkill
- regex-vs-llm-structured-textSkill
- rules-distillSkill
- scrapling-officialSkill
- session-auditSkill
- shinen-designSkill
- skill-stocktakeSkill
- token-budgetSkill
- vault-indexSkill
- vault-ingestSkill
- vault-lintSkill
- vault-querySkill
The file
We split this skill into the rules it gives an agent
2 rules. About 0 are common, 1 is distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈1
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 43
Do NOT vendor the NVIDIA repo into this plugin — it is large and Apache-2.0 licensed.
- Very specific: line 88
Use this on anything pulled from a third-party marketplace before it lands in ~/.claude/skills or ~/.claude/plugins.
≈ “Common” is an estimate: at least 1 in 5 of the 350,495 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
SkillSpector — Agent Skill Security Scanner
Answers one question: "Is this skill safe to install?" Research behind the tool found 26.1% of public skills contain vulnerabilities and 5.2% show likely malicious intent. Scan before you trust.
Activation
User says something like:
- "scan this skill: <path|url|repo>"
- "is this plugin safe to install?"
- "audit
<skill>for security" - "vet this MCP before I add it"
- "security-check my skill before publishing"
Ensure the tool is installed
skillspector is a standalone CLI (Python 3.12+). If which skillspector is empty,
install it once — prefer uv:
uv tool install git+https://github.com/NVIDIA/skillspector.git
# update later: uv tool update skillspectorDo NOT vendor the NVIDIA repo into this plugin — it is large and Apache-2.0 licensed. This skill only orchestrates the installed CLI.
Run a scan
Input can be a directory, single file, git repo URL, remote URL, or zip.
# Fast static-only scan (no API key needed) — default for quick triage
skillspector scan <target> --no-llm
# Full scan with LLM semantic pass (filters false positives, explains findings)
SKILLSPECTOR_PROVIDER=anthropic ANTHROPIC_API_KEY="$ANTHROPIC_API_KEY" \
skillspector scan <target>
# Machine-readable report to a file
skillspector scan <target> --no-llm --format json --output report.json
# formats: terminal (default), json, markdown, sarifInterpreting output
- Risk score 0–100 with severity label + recommendation. Higher = more dangerous.
- Findings are categorised (e.g. SC4 = vulnerable dependency via OSV.dev, prompt injection, data exfiltration, MCP tool poisoning). Read the category + the matched snippet, don't just trust the number.
- Static stage = regex/AST/taint/YARA (fast, can false-positive). The LLM stage prunes false positives and explains intent — run it for anything you're about to actually install.
Suppressing known false positives
Re-scans only need to surface new issues. Accept known findings with a baseline:
skillspector scan <target> --no-llm --baseline .skillspector-baseline.jsonReport back
Summarise: risk score, the top high/critical findings (category + file:line + why), and a clear verdict — safe / review-needed / do-not-install. For audits before publishing, list each finding the author should fix.
Note
Use this on anything pulled from a third-party marketplace before it lands in
~/.claude/skills or ~/.claude/plugins. The local install lives at
~/.local/bin/skillspector.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.