taste
Pass
Audited by Gen Agent Trust Hub on Sep 12, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill acts as an orchestrator for multiple video production tools (FFmpeg, Remotion, Fal.ai) and ingests external project notes to generate shot plans and prompts, creating a surface for indirect instructions. The project notes are used to map aesthetics and sequence shots, which could be exploited to inject instructions into downstream generation and composition steps.
- Ingestion points: External 'project notes' referenced in the 'Beat-Mapped Shot Plan' section are used to define the production logic.
- Capability inventory: The skill executes shell commands via FFmpeg and
npx remotion, and interacts with external image/video generation APIs. - Boundary markers: The instructions lack explicit delimiters or 'ignore' directives to separate user project data from agent instructions.
- Sanitization: No sanitization or validation logic is specified for external inputs before they are interpolated into prompt templates or shell commands.
- [COMMAND_EXECUTION]: The skill provides specific FFmpeg recipes and Remotion render commands to be executed in the terminal for video processing, cropping, and composition.
- Evidence: Multiple FFmpeg command snippets and an 'npx remotion render' command are provided in the 'FFmpeg Recipes' and 'Remotion Composition Skeleton' sections of
SKILL.md. - [DYNAMIC_EXECUTION]: The skill generates a React-based composition script (Remotion) at runtime by adapting a provided skeleton with beat-synced sequences and dynamic asset paths.
- Evidence: The 'Remotion Composition Skeleton' section provides a TypeScript React template intended for dynamic population and execution via the Remotion CLI.
Audit Metadata