gepetto

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

• Third-party content exposure detected (high risk: 1.00). The skill's research step (references/research-protocol.md and SKILL.md Step 5) explicitly performs WebSearch/WebFetch on open web results and ingests those third‑party sources into claude-research.md, which the agent then uses to drive interviews, spec synthesis, and planning—clearly exposing the agent to untrusted public content that can influence subsequent actions.