NVIDIA open-sources SkillSpector: 26% of AI agent skills contain vulnerabilities
blaizedsouza · x · 2026-11-14
NVIDIA released SkillSpector, a free, open-source tool for scanning AI agent skills for security risks before they enter your workflow.
Research cited in its README is striking: of 31,132 skills analyzed, 26.1% contained at least one vulnerability and 5.2% showed patterns suggesting malicious intent.
Basic usage:
bash
skillspector scan ./my-skill/ --no-llm
That runs a static scan; you then review findings alongside a risk score. A low score is useful evidence, but not proof a skill is safe. Fully free and open-source.
More from coding & agent
- Open-source 3D globe with live flight data, satellite imagery and MCP support (64K+ stars) — iamfakhrealam · 2026-11-14
- Claude Max hailed as 2026's best investment as Fable 'spawned thousands of startups' — EXM7777 · 2026-11-14
- QuixiAI's Rust tool adds local model support via Ollama, llama.cpp with cloud fallback — QuixiAI · 2026-11-14
- ARC 1.0: A 0.6B Apache 2.0 Model That Routes AI Agents Across Cloud and Local Models — Balance- · 2026-11-14
- DevDay swag turned into Open.Dotlet, a 30-key physical prompting tablet for Codex — paw_lean · 2026-11-14
- Helium Solo: Free Tool to Record, Replay and Debug Failed AI Agents Locally — kasankilabs · 2026-11-14