fermi-estimation

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

• Third-party content exposure detected (high risk: 0.80). SKILL.md (Step 2: "Evidence First") explicitly instructs the agent to "autonomously gather accessible evidence," "prefer current primary sources," and "use available tools and public sources proactively" (e.g., product/pricing pages, industry reports), which requires fetching and interpreting open/public third‑party content and thus exposes the agent to untrusted web content that could carry indirect prompt injection.