build-daily-inspiration-sites

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

• Third-party content exposure detected (high risk: 0.75). SKILL.md step 1 runs scripts/validate_capture.mjs and step 3/4 then reads manifest.json, content.md, and attaches fullPageImage as evidence for each item; since these capture artifacts are supplied by “the article path supplied by the user” (outsider-authored capture files), the agent’s LLM context can receive untrusted free-form text from content.md / manifest-derived prompts via the task briefs.