reverse-information-paradox

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

• Third-party content exposure detected (medium risk: 0.30). SKILL.md’s runtime workflow (“Coach mode” step 3 and “The Process” step 1) elicits and audits the user’s real case content about proprietary prompts/tickets/corrections/evals and then has the agent ingest and reason over that outsider-authored free text from the user to produce the Intelligence Exhaust Map and trust-boundary decisions.