NVIDIA/SkillSpector

SkillSpector is NVIDIA's open-source security scanner for AI agent skills — the folders of instructions and scripts used by tools like Claude Code, Codex CLI, and Gemini CLI — built to answer whether a given skill is safe to install before running it with an agent's implicit trust. NVIDIA cites research finding that 26.1% of agent skills contain vulnerabilities and 5.2% show likely malicious intent, and SkillSpector checks for 68 vulnerability patterns across 17 categories including prompt injection, data exfiltration, privilege escalation, supply-chain risk, and MCP tool poisoning. It can scan a skill from a Git repo, URL, zip file, directory, or single file, combining fast static analysis with an optional LLM-based semantic evaluation, and cross-references OSV.dev for real-time CVE data. Results come with a 0-100 risk score, severity labels, and reports in terminal, JSON, Markdown, or SARIF format, with a baseline/suppression system so re-scans only surface new findings.