pr-comments

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

• Third-party content exposure detected (high risk: 1.00). The skill explicitly fetches GitHub PR review comments using the REST/GraphQL endpoints (e.g., gh api repos/{owner}/{repo}/pulls/{pr_number}/comments and reviewThreads) — untrusted, user-generated third-party content the agent reads and uses to decide edits, apply suggestions, and take actions, which can enable indirect prompt injection.

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

• Potentially malicious external URL detected (high risk: 0.70). The skill fetches untrusted review comment bodies at runtime from the GitHub REST/GraphQL endpoints (e.g., gh api repos/{owner}/{repo}/pulls/{pr_number}/comments and gh api graphql) which can contain instructions or suggestion blocks that directly influence the agent's edits/behavior, so these URLs are a runtime source that could control prompts or actions.