lov-gh-contribute

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

• Third-party content exposure detected (medium risk: 0.30). The required workflow reads upstream GitHub repository files like CONTRIBUTING.md and CODE_OF_CONDUCT.md via gh api repos/$UPSTREAM/contents/... and also may use .github/PULL_REQUEST_TEMPLATE.md text, where these files can be outsider-authored on public repos (e.g., via PR/commit history) and are ingested into the runtime.

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

• Potentially malicious external URL detected (high risk: 0.90). The skill issues a runtime gh API call to fetch CONTRIBUTING.md (gh api repos/$UPSTREAM/contents/CONTRIBUTING.md -q .content 2>/dev/null | base64 -d) and uses that fetched content (and PR templates) to shape commit/PR instructions and templates, so remote content directly controls agent behavior.