Social Corpus Harvest

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

• Third-party content exposure detected (low risk: 0.10). The required workflow fetches posts from user-selected Xiaohongshu, Douyin, and Weibo profile links through OpenCLI, so outsider-authored text is ingested only after the agent actively fetches specific profiles, not from a feed an outsider can post into.

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

• Potentially malicious external URL detected (medium risk: 0.30). The instructed npx skills add setup fetches skill material from the individually owned GitHub repository yuwanpai2004-create/codex-skills, which is unofficial with no enforced digest or signature, and no concrete warning sign is shown, so the tier is medium.