pov-vlog-presence-by-1hv6g2b
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from local and external sources. Ingestion points: Data enters through the references/cases.md file and external links to MP4 and JPG media. Boundary markers: While the skill uses a 'Preserve / Replace / Avoid' reference contract to structure its output, it lacks explicit security delimiters or 'ignore' instructions for the external content. Capability inventory: The skill invokes browser and vision tools to inspect the finished media of anchor cases. Sanitization: No explicit sanitization or filtering mechanisms for potential instructions embedded in the external prompts or media metadata are implemented.
Audit Metadata