teach

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

• Third-party content exposure detected (high risk: 0.70). SKILL.md instructs the agent to read workspace files like ./reference/*.html, MISSION.md, RESOURCES.md, and ./learning-records/*.md—which are free-text documents sourced from files the user/outer system provides—so at runtime the LLM can ingest outsider-authored content (including any HTML/markdown the operating user didn’t write themselves).