panel-review

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md resolves the target by reading GitHub PR JSON (title/body via gh pr view ... --json title,body,...) and then passes the PR’s args.claim/diff scope into lens and verifier sub-agents, so an outsider-authored PR body is ingested as free text without requiring selecting a specific item first.