social-auto-publisher
Pass
Audited by Gen Agent Trust Hub on Aug 8, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [EXTERNAL_DOWNLOADS]: The script
scripts/generate_content.jsmakes a network request to an external domainhttps://newsnow.busiyi.world/api/hottestto fetch trending topics. This domain is not on the standard whitelist, though the action is aligned with the skill's purpose of tracking trends. - [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection attack surface.
- Ingestion points: The
scripts/monitor_interactions.jsscript fetches untrusted data (comments, mentions, and messages) from Xiaohongshu, Weibo, and Twitter platforms. - Boundary markers: There are no explicit boundary markers or instructions to the model to ignore or safely handle embedded commands within the ingested social media content.
- Capability inventory: The skill has the capability to perform network operations via APIs and browser automation to post content and replies (
scripts/publish.js,scripts/auto_reply.js,scripts/utils/xiaohongshu.js). - Sanitization: The skill performs minimal sanitization of ingested content (primarily regex-based cleaning for hashtags) before processing it for classification or using it as context for automated replies.
Audit Metadata