media-composition
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-provided manifest.yaml files. The manifest schema includes a 'requires' field for commands to run before generation, creating a surface for indirect injection. Ingestion points: manifest.yaml. Boundary markers: Absent. Capability inventory: ffmpeg, yq, and bash shell. Sanitization: Absent.
- [COMMAND_EXECUTION]: Shell commands use variable interpolation from manifest fields. The pattern 'for output in $(yq eval ...)' and ffmpeg arguments lack sanitization against shell metacharacters.
Audit Metadata