run-corpus-research

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

• Third-party content exposure detected (high risk: 1.00). The skill’s required run-research integration block (references/subagent-briefs.md) and the orchestration/tool-steering guidance (references/orchestration.md and SKILL.md) explicitly instruct subagents to perform raw/smart web searches and raw-scrape-links on public websites and Reddit, which the agents must read and synthesize into decision-grade outputs—i.e., untrusted, user-generated third‑party content is fetched and used to drive decisions.