outlier-post-finder
Pass
Audited by Gen Agent Trust Hub on Nov 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data from social media platforms, which could contain malicious prompts meant to hijack agent logic. Ingestion points: External data such as video transcripts, post captions, and comments are fetched from various social media feeds as described in SKILL.md. Boundary markers: Instructions in the Pitfalls section explicitly state that captions and comments are untrusted and any instructions inside them should be ignored. Capability inventory: The skill focuses on data aggregation, median calculation, and pattern explanation; it does not request high-risk permissions like shell access or file system persistence. Sanitization: The skill relies on natural language instructions to separate data from instructions during analysis.
Audit Metadata