skill-ugc-discovery
Pass
Audited by Gen Agent Trust Hub on Sep 2, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill represents an attack surface for indirect prompt injection as it is designed to ingest and process untrusted content from external social media platforms (Xiaohongshu, Bilibili, Weibo, etc.) using
WebSearchandWebFetchtools. - Ingestion points: Untrusted content is retrieved from multiple social media platforms via web search results and full-page fetches (SKILL.md, Execution Steps 3 and 4).
- Boundary markers: The skill uses a structured markdown report template which provides some structural separation, though it does not explicitly instruct the agent to ignore or escape instructions contained within the fetched content.
- Capability inventory: The skill utilizes web search and retrieval tools, along with the ability to generate and save structured reports to the
outputs/directory. - Sanitization: No specific sanitization or filtering logic is defined for the fetched external content beyond sentimental classification.
- [EXTERNAL_DOWNLOADS]: The skill documentation in
EASEL-META.mdreferences an external GitHub repository,NanmiCoder/MediaCrawler, as a technical reference for its crawling logic. This is an informative reference and does not involve automated downloads or execution during the skill's operation.
Audit Metadata