Observed arrival · 2026-11-05
Skillverdict: a scorecard for the skill scanners
A benchmark of four open-source AI-agent skill security scanners, tested against 186 skills in four languages and a set of attack samples.
○Open
⊠Login
$Paid
†Ads
✦Pretty
●Pro
◎Niche
◉Human
⚑Risk
ƒJS
- For
- People evaluating security scanners for AI-agent skills
- Worth noticing
- The project reports a 0.71 rank correlation between code-file count and SkillSpector’s score; 38 of 43 legitimate skills shipping code got “DO NOT INSTALL.”
Field notes
The page says it ran four open-source static scanners and its own language-aware LLM review against 186 skills in English, Russian, German, and Chinese, plus 40 defanged attack samples. It reports a 0.71 rank correlation between code-file count and SkillSpector’s score, and says none of 43 legitimate skills with six or more code files received SAFE. The site also links to a case study about Anthropic’s claude-api skill.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
○OpenPublic substance visible
✦PrettyNotable craft visible
●ProPolished or operationally mature
◎NicheUnusually specific use
One card from the complete issue