make-montage
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes external video and audio files ($0 and $1) which serve as ingestion points for untrusted data. These files are analyzed by the Gemini model during the montage creation process. Maliciously crafted metadata or content within these files could potentially influence the model's creative judgment or reasoning.
- Ingestion points: External media files provided as command-line arguments.
- Boundary markers: The instructions lack boundary markers or warnings to ignore instructions embedded in media metadata.
- Capability inventory: The skill has file system write capabilities via the 'render-final' tool and can modify project timelines.
- Sanitization: No sanitization or validation of input file metadata is performed before the model analysis steps.
Audit Metadata