mj-writer

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

• Third-party content exposure detected (high risk: 0.90). Yes — the SKILL.md and README explicitly say the agent ingests user-provided "links" and materials (PDF/语音转文字) and references/content_methodology.md lists public social sources (Twitter, Reddit, 小红书, 微博等) as selection sources, so the agent is expected to read and use untrusted, user-generated third‑party content to drive writing decisions and outputs.