known-issues

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

• Third-party content exposure detected (medium risk: 0.30). The skill’s required runtime workflows for search, check-all, and quality invoke gh issue view / gh search issues to ingest outsider-authored free text from GitHub issue bodies/titles/comments at runtime (e.g., those issue fields are read and then used for triage/registry updates).

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

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly fetches live GitHub issue templates at runtime (e.g. via "gh api repos/anthropics/claude-code/contents/.github/ISSUE_TEMPLATE/.yml --jq '.content' | base64 -d" / "gh api repos/{repo}/contents/.github/ISSUE_TEMPLATE/{template_file} --jq '.content' | base64 -d"), and those fetched templates directly control how the agent drafts issue bodies (i.e., they control prompts/instructions), so this is a required runtime external dependency.