pov-vlog-presence-by-ha1tt9
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill presents a surface for indirect prompt injection by analyzing external media and reference data as part of its prompt synthesis workflow.
- Ingestion points: The skill reads local reference data from
references/cases.mdand is instructed to analyze external assets fromgoodcase.aiandx.com. - Boundary markers: The workflow incorporates a 'Preserve / Replace / Avoid' reference contract to manage data influence, though it lacks explicit safety instructions to disregard commands hidden within media content.
- Capability inventory: Instructions in
SKILL.mdauthorize the agent to use image, video, and browser tools to inspect external reference materials. - Sanitization: No explicit methods for sanitizing or filtering instructions from external media or documentation are defined.
- [EXTERNAL_DOWNLOADS]: The skill references external URLs from
goodcase.aiandx.comto provide the necessary context and media examples for its video prompt generation logic.
Audit Metadata