remotion
Pass
Audited by Gen Agent Trust Hub on Jun 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes shell commands to manage project lifecycles, including initializing Remotion projects with
npm create video, installing necessary dependencies, and rendering the final video output vianpx remotion render. It also includes a custom Bash script for asset retrieval usingcurl. - [EXTERNAL_DOWNLOADS]: Fetches design screenshots and HTML metadata from Google Cloud Storage via the Stitch platform. It also downloads standard video animation libraries from the npm registry to support the rendering environment.
- [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and processes untrusted data (titles, descriptions, and HTML) from external Stitch projects. If this data contains malicious instructions, it could influence the agent's behavior during the video generation phase.
- Ingestion points: Screen titles, descriptions, and HTML content retrieved via
stitch:list_screensandstitch:get_screen(SKILL.md). - Boundary markers: Absent; there are no explicit delimiters or instructions provided to the agent to treat design metadata as untrusted data during interpolation.
- Capability inventory: Subprocess execution (
npm,npx), file-system modification (Write), and network operations (web_fetch,curl) across several scripts. - Sanitization: Absent; the skill does not implement validation or escaping for design content before it is used to generate React components.
- [COMMAND_EXECUTION]: Implements dynamic code execution by generating React/TypeScript components (
WalkthroughComposition.tsx,ScreenSlide.tsx) at runtime based on design metadata. These generated scripts are subsequently executed by the Remotion engine to produce the final MP4 file.
Audit Metadata