wechat-draft-publish

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

• Third-party content exposure detected (medium risk: 0.30). Skill runtime path scripts/preprocess-and-publish.mjs: it reads user-provided free text --html/--content-file and scans <img src> values, then base64-encodes and uploads local file bytes derived from that HTML (and also includes --title/--digest), so outsider-authored HTML is ingested at runtime.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.70). The skill documentation instructs agents to fetch runtime writing instructions from https://doubaoya.com/api/wechat/writing-spec and "照它写即可", meaning the fetched markdown would directly control the agent's generation prompts at runtime, so this external URL is a runtime dependency that controls prompts.