table-video

Pass

Audited by Gen Agent Trust Hub on Nov 18, 2026

Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses child_process.spawn and execFile to run ffmpeg and ffprobe for media analysis and extraction.
  • Evidence found in scripts/motion-check.mjs, scripts/stillness.mjs, and scripts/preflight.mjs.
  • These operations are limited to specific media processing tasks and include timeouts and buffer limits.
  • [DYNAMIC_EXECUTION]: The skill dynamically loads Node.js modules (e.g., Remotion, React) from user-specified paths provided via the --runtime argument.
  • Evidence found in scripts/render-motion-example.mjs and assets/motion/production-core/render.cjs.
  • This mechanism is necessary for the tool to function within the user's existing Remotion development environment.
  • [EXTERNAL_DOWNLOADS]: The skill includes a utility to fetch reference images from remote URLs.
  • Evidence found in scripts/save-reference-image.mjs.
  • The implementation includes several security measures: it scrubs credentials/queries from URLs, validates image headers (magic bytes) to ensure they are real images, enforces byte size and pixel dimension limits, and prevents path traversal during storage.
  • [INDIRECT_PROMPT_INJECTION]: The skill possesses a significant attack surface as it processes untrusted user-supplied content, such as narration scripts, motion graphic definitions, and external project files.
  • Ingestion points: Files in 01_稿件/, 02_配音/, 03_视觉设计/, and 05_制作工程/.
  • Boundary markers: The skill uses structured JSON schemas and Markdown sections (e.g., md:D01) to delimit content.
  • Capability inventory: Includes subprocess spawning (ffmpeg), network fetching (images), and extensive local file write capabilities.
  • Sanitization: Includes internal image header validation, URL sanitization, and strict path verification to prevent project escape.
Audit Metadata
Risk Level
SAFE
Analyzed
Nov 18, 2026, 02:58 AM