create-code

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

• Third-party content exposure detected (high risk: 0.80). The orchestrator explicitly calls the external dogpile search tool (see stage_2_research in orchestrator.py and SKILL.md: "Calls /dogpile search ...") and captures its output into context['research_summary'], which means untrusted/public web/search results are ingested and used to influence later workflow and decisions.