decompose

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

• Third-party content exposure detected (medium risk: 0.65). At runtime the skill reads docs/plans/<workflow_id>.md (Step 2/3/4) and ingests the fenced ## Slices YAML authored by whoever created the plan doc; since that plan doc is not guaranteed to be operating-user-authored, it can contain outsider-authored free text that becomes LLM context.