search-skill

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). SKILL.md’s runtime workflow for “search-skill” ingests and treats as untrusted free-text candidate descriptions fetched from external catalogs (e.g., via npx skills find and WebFetch https://skills.sh/search?...) before scoring and potentially installing an outsider-authored skill.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The skill explicitly performs runtime network fetches and installs third-party skill tarballs (bin/add-skill uses curl / may WebFetch tarball URLs), for example via the skills.sh search URL https://skills.sh/search?q= and curated repo references like vercel-labs/agent-skills, which means remote code will be fetched and installed at runtime.