horror-suspense-short-by-43hd7f
Pass
Audited by Gen Agent Trust Hub on Oct 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill defines a structured motion-briefing workflow that relies on documented case studies. It uses standard inputs for video generation (duration, aspect ratio, model) and adheres to safe attribution practices.
- [EXTERNAL_DOWNLOADS]: The skill references external URLs from
goodcase.aiandx.comwithinreferences/cases.md. These links serve as legitimate source material and reference media for the prompt engineering workflow and are displayed in plain text without obfuscation. - [INDIRECT_PROMPT_INJECTION]: The skill exhibits an inherent attack surface for indirect prompt injection by processing external prompt excerpts and user-supplied story details. This is necessary for its primary function of synthesizing new prompts from evidence-based cases.
- Ingestion points: User inputs defined in
SKILL.mdand external reference data inreferences/cases.md. - Boundary markers: The skill uses a 'Preserve/Replace/Avoid' reference contract to structure the output, which provides logical boundaries, though it does not explicitly use escaping for external content.
- Capability inventory: The skill uses prompt generation logic and requests the use of media/browser tools to inspect reference content.
- Sanitization: None detected beyond standard instructional constraints.
Audit Metadata