the-one-thing

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

• Third-party content exposure detected (high risk: 0.80). The SKILL.md "Process — putting it to work" step 4 explicitly requires the agent to "Research benchmark + trend before answering (p. 112)" and even says "Books and the internet; find who's studied or accomplished this," which mandates fetching and interpreting open/public third‑party web content (untrusted user/website sources) that can materially influence the agent's decisions and next actions.