init-project-research

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

• Third-party content exposure detected (high risk: 0.80). The skill explicitly runs an automated literature & discovery phase (Phase 9) that spawns a sub-agent to query open/public sources ("Sources: OpenAlex + Scopus" in Phase 9 / references/literature-discovery.md) and a "scout novelty" tool to ingest results, and then reads and integrates those external results into the Atlas topic file and project decisions (e.g., flagging low novelty), so untrusted third‑party content is fetched and can materially influence subsequent actions.