extract-signal

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md describes an agentic workflow that uses gather tools like /prime-sweep and /prime-fetch to pull from “many sources”/“one noisy call,” which at runtime can include outsider-authored free text from public web content or other third-party messages that the sub-agent reads into its LLM context.