Skip to content

Security & trust

Skillboard indexes attacker-controlled repositories. We treat all repository text as hostile input and never execute it.

What we do

  • Fetch only through the GitHub API; never run repository scripts, hooks or package managers.
  • Parse text with strict size, depth and alias limits, and disable raw HTML in rendered Markdown.
  • Run static rules that detect and explain risky patterns, with rule and scanner versions recorded.
  • Pin every listing to an exact commit SHA and content hash.

What static checks cannot do

A static check is not proof of safety. It cannot observe runtime behavior, and a skill with no findings may still be harmful. We never label a skill simply “safe”. You are responsible for reviewing the source and the exact diff before running anything. See our methodology.

Licensing and opt-out

Public visibility on GitHub does not grant redistribution rights. We show source attribution and recognized license status, and display the full body only when the license permits – otherwise a bounded excerpt with a link to the source. Maintainers can request removal or opt-out.

Reporting

To report a mislabeled, malicious or opted-out skill, contact the maintainers of Skillboard through the project's source repository. A structured in-app report flow is planned.