implement

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md requires reading “the spec or ticket” (outsider-authored free text) before launching subagents, which would be ingested into the agent’s LLM context; however, the workflow text itself also doesn’t show the exact runtime retrieval mechanism, so exposure is plausible but not fully evidenced.