blog-writer

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

• Third-party content exposure detected (high risk: 0.90). The skill's required research workflow and references explicitly direct the agent to perform external web research and read public pages (e.g., references/ideation-and-research.md: "search the target keyword, read the top 10 results" and "Targeted external research — ... authoritative websites") and the distribution/reference docs name public platforms (Quora, Reddit, SlideShare, YouTube), so the agent is expected to ingest and act on untrusted, user-generated or public web content as part of its pipeline.