async-learning-teacher

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

• Third-party content exposure detected (medium risk: 0.30). In SKILL.md “Workflow” step 1–2, the agent parses user-provided captures (including “Pasted source text” in step 1 and, if access is unavailable, uses “the URL, title, abstract, snippet, or pasted user context” in step 2), so outsider-authored free text can be ingested directly into the LLM without selecting a specific trusted item.