account-research

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

• Third-party content exposure detected (high risk: 0.85). The skill’s runtime LLM context is populated by the Extruct Deep Research report returned from POST /v1/deep_research_tasks (including report.markdown and report.sources), which is compiled from public web content and other third-party sources; those cited pages’ text is ingested into the agent’s context as free-form prose during dossier assembly.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill explicitly invokes the Extruct Deep Research API at https://api.extruct.ai/v1/deep_research_tasks (POST to create tasks and GET to poll results) and consumes the returned report.markdown/report.sources at runtime to drive the dossier assembly, so external content from that URL directly controls the agent's output and is a required runtime dependency.