self-learning

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

• Third-party content exposure detected (high risk: 0.90). Yes—SKILL.md step 3 (“URL Reading”): it selects URLs found via runtime web search and then uses read_url_content / browser_subagent to extract main page content as markdown, which is then fed into the agent during “Generate Skill” (prompt/content synthesis), making outsider web-authored free text an LLM context input.