Agent skills & prompts
NVIDIA/SkillSpector
Scan agent skills for risks before installing
- 15k
- Apache-2.0
- Python
What it is and why it matters
NVIDIA's security scanner that analyzes agent skills (Claude Code, Codex, MCP) for malicious patterns, data exfiltration, prompt injection and supply-chain risks. It ships several scanners and ranks each skill by risk level, pointing to file and line. Useful when installing third-party skills whose code, sometimes thousands of files, you will not review by hand.
Key points
- Detects data exfiltration, prompt injection and suspicious patterns
- Ships deliberately dangerous test skills to verify the scanner
- Recommended for marketplace operators accepting user-uploaded skills
Where we saw it
The Next New Thing 2026-06-19 The hosts describe it as antivirus for skills; they note it reports the problem file and line, and suggest running any downloaded skill through it. Free CapCut, marketing tools, and all of github’s top 10 repos The Next New Thing 2026-07-24 The guest values the NVIDIA name behind it and advises scanning a skill before using it, even recreating it after reviewing what the scan flags. Free ai video creation + OCR full books + private photos + more Github finds
Same category