deep-research

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md workflow explicitly requires fetching external web content — e.g., "Step 2: Fetch Top Sources" instructs to "Run WebFetch with the URL" (and loads source map from .claude/tmp/_sources-{slug}.md produced by web-research), so the agent ingests and acts on arbitrary public web pages that could contain untrusted instructions.