codebase-5yr-forecast

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

• Third-party content exposure detected (high risk: 0.85). Phase 3 “External Evidence Gathering” uses web-search/web-reader to fetch and read public web pages at runtime, and that retrieved free-form text is then used by the LLM to fill scoring_input.json (LLM context), creating an indirect prompt-injection risk.