NVIDIA open-sources SkillSpector to scan AI agent skills for malicious code
dr_cintas · x · 2026-07-23
- NVIDIA’s SkillSpector is a free security scanner for AI agent skills used by tools like Claude Code, Codex CLI, and Gemini CLI.
- The pitch is simple: skills run with broad access, so a convenient skill can also leak secrets or introduce malicious behavior.
- SkillSpector can scan Git repos, URLs, zip files, directories, or single files, then produces a risk score from 0–100 with a verdict such as safe, caution, or do not install.
- It combines fast static analysis with an optional LLM-based semantic pass, checks live vulnerabilities via OSV.dev, and exports terminal/JSON/Markdown/SARIF reports.
- The repo claims coverage of 68 vulnerability patterns across 17 categories, including prompt injection, data exfiltration, privilege escalation, system prompt leakage, memory poisoning, MCP tool poisoning, and dangerous code.
- The screenshot also cites research saying 26.1% of skills contain vulnerabilities and 5.2% show likely malicious intent.
Related event: NVIDIA Open-Sources SkillSpector for AI Agent Security(3 posts)→
More from coding & agent
- Drex Search cuts agent steps 62% and tokens 56% with decision-powered retrieval — _AustinCalvert_ · 2026-10-27
- Engineer's real numbers: 20B tokens in 4 months, agents still beat hiring by a mile — thechrisperry · 2026-10-27
- Vovy launches a cursor-following voice agent for Mac, free tier gives 100 credits/mo — Scobleizer · 2026-10-27
- Redditor seeks real-world stacks for a persistent-memory personal AI that answers texts and emails — BrooksomidusGG · 2026-10-27
- Practical tip: use max effort for planning, xhigh for implementation in Claude Code — majidmanzarpour · 2026-10-27
- Chrome team explains WebMCP: agents get tools from your website, not a replacement for MCP — rseroter · 2026-10-27