Security audit of NVIDIA/SkillEvaluator · Codex Skill by NVIDIA · ★ 4.0k
Yes — SkillEvaluator passed AgentSkillsHub's rule-based security scan with no dangerous patterns detected. As with any third-party skill, confirm what credentials it requests before production use.
What it is: Multi-tier framework for evaluating AI agent skills and plugins with quality gates, security scanning, harness compatibility checks, semantic overlap detection, synthetic evaluation dataset generation, and live agent evaluation that measures how skills affect agent behavior. Ships reusable CI templates that bring every check to any repository.
No dangerous patterns were detected: no credential exfiltration, no obfuscated downloads, no sandbox-escape attempts, no prompt-injection markers.
| Security grade | ✓ SAFE |
| Quality score | 69/100 |
| GitHub stars | 4.0k |
| Language | Python |
| License | Apache-2.0 |
| Last updated |
Browse all Codex Skill skills →
This is AgentSkillsHub's free basic audit: an automated rule-based scan with 44 published rules built on SlowMist's agent-security taxonomy (credential theft, obfuscated payloads, privilege escalation, covert instructions to the agent, and more), applied the same way to every open-source AI agent skill and MCP server we grade and refreshed every 8 hours. A SAFE grade is a scan result, not a guarantee — deep 5-dimension audits (code · credentials · vendor · supply-chain · operational) are available for enterprise. Audited: .