implement-plan-with-subagents

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

• Third-party content exposure detected (high risk: 0.78). The workflow reads outsider-authored free text into LLM context via user-provided plan artifacts: it reads docs/threads/<...>/plan.md and plan-tasks/*.md (which may include an outsider-named Source:) and then passes those contents/derived “Assumptions / Known risks” into dispatched subagents’ briefs and re-reviews (step 4 + step 6b/6c), creating an indirect prompt-injection exposure path.