nemoclaw-maintainer-e2e

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md dispatches and then ingests GitHub Actions API JSON (run/job metadata) via gh api ... repos/NVIDIA/NemoClaw/actions/runs/$RUN_ID and runs validate-full-e2e-evidence.mts, which parses dispatch.json, run-json, jobs-json, and launchable-e2e-json at runtime; those API fields can ultimately include outsider-authored free text from the triggered workflow’s repository context (e.g., commit messages/test logs) even though the skill controls selectors.