frame-and-explain

Pass

Audited by Gen Agent Trust Hub on Jul 26, 2026

Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDATA_EXFILTRATION
Full Analysis
  • [COMMAND_EXECUTION]: The skill provides numerous instructions for executing shell commands to manage a Remotion project.
  • Examples include scaffolding new projects with npx create-video, adding official plugins via npx remotion add, and rendering videos using npx remotion render.
  • It also includes instructions for running custom Node.js scripts (e.g., generate-voiceover.ts) and invoking FFmpeg for video processing.
  • [EXTERNAL_DOWNLOADS]: The skill references several external resources for assets and configuration.
  • Fetches sound effects, videos, and documentation from remotion.media and remotion.dev.
  • Downloads Lottie animation files from assets4.lottiefiles.com and MapLibre styles from demotiles.maplibre.org.
  • Downloads the Whisper.cpp binary and models via the @remotion/install-whisper-cpp package for audio transcription.
  • [DATA_EXFILTRATION]: To support AI-generated voiceovers, the skill provides patterns for sending text data to api.elevenlabs.io.
  • This process requires the use of an ELEVENLABS_API_KEY, which the skill correctly instructs to be handled via environment variables rather than being hardcoded.
  • [INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests untrusted data that influences the generation of video content.
  • Ingestion points: calculateMetadata fetches JSON data from external URLs; defaultProps in Composition objects; and external SRT/JSON caption files.
  • Boundary markers: None explicitly defined in the provided code snippets.
  • Capability inventory: The skill can write files to the local file system (fs.writeFileSync), perform network requests (fetch), and execute shell commands (ffmpeg).
  • Sanitization: No specific sanitization or validation of the fetched content is demonstrated in the examples.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 26, 2026, 01:17 AM
Security Audit — agent-trust-hub — frame-and-explain