travel-city-walk
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [SAFE]: The skill provides a structured creative workflow utilizing a local reference file (references/cases.md) to guide the generation of video prompts.
- [SAFE]: Explicit safety constraints are included to prevent the agent from performing unauthorized account changes, purchases, or private asset uploads.
- [INDIRECT_PROMPT_INJECTION]: The skill processes prompt excerpts from external creators to facilitate style transfer. 1. Ingestion points: Prompt excerpts located in references/cases.md. 2. Boundary markers: The 'Preserve / Replace / Avoid' reference contract requirement acts as a logical filter and boundary. 3. Capability inventory: The skill is restricted to text artifact generation; it lacks subprocess calls, file-writing, or network exfiltration capabilities. 4. Sanitization: Procedural comparison with the anchor case and revision steps are required to ensure the output aligns with creative intent.
- [EXTERNAL_DOWNLOADS]: The skill references external media assets and case studies on goodcase.ai and established social media platforms for visual grounding. These references are used for analysis and do not involve executable code or sensitive system access.
Audit Metadata