double-check

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches and reads GitHub PR review comments and reviews via gh pr view ... --json comments --jq '.comments[].body' and --json reviews --jq '.reviews[].body' (Step 1 / "Reviewer reads CI comments directly via gh"), which are untrusted, user-generated third-party content that the agent is required to interpret and that can change its curation, fixes, and subsequent actions.

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

• Potentially malicious external URL detected (high risk: 0.80). The skill clones the PR repository at runtime using "https://github.com/$REPO.git" and then runs the repo's tests/commands (e.g., npm test, pytest), which means fetched remote code from that URL is executed as part of the workflow.