skills/involvex/skills/remotion/Gen Agent Trust Hub

remotion

Pass

Audited by Gen Agent Trust Hub on Apr 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill downloads screenshots and HTML metadata from the Stitch platform (typically hosted on Google Cloud Storage) to serve as visual assets for the generated video. These downloads are initiated via the web_fetch tool or a dedicated bash script using curl.
  • [COMMAND_EXECUTION]: The skill uses the shell to manage a Node.js environment, including initializing projects with npm create, installing official Remotion packages, and executing the remotion render command to produce the final MP4 output.
  • [REMOTE_CODE_EXECUTION]: While the skill generates React/TypeScript code at runtime, this code is constructed from local templates and specific design data fetched from Stitch. The execution occurs within the local Remotion rendering environment as part of the intended video creation workflow.
  • [DATA_EXFILTRATION]: The skill accesses design data from Stitch projects but does not exhibit patterns of sending sensitive information to unauthorized third-party domains. All processing and video rendering are performed within the agent's project environment.
  • [PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface because it ingests and processes untrusted data (screen titles, descriptions, and HTML content) from Stitch projects. Malicious instructions embedded in these fields could theoretically attempt to influence the agent's behavior during the composition phase, although the impact is limited by the skill's specific task scope.
Audit Metadata
Risk Level
SAFE
Analyzed
Apr 16, 2026, 09:57 PM