editing-videos
Pass
Audited by Gen Agent Trust Hub on Oct 4, 2026
Risk Level: SAFECOMMAND_EXECUTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: Multiple templates and helper scripts, including edit-engine.mjs, record.example.mjs, and gen-voice.mjs, use child_process.execFileSync to run ffmpeg, ffprobe, and playwright. These are used to record screen sessions, extract frames, measure audio loudness, and transcode video files. Additionally, the Python script library/tools/synthesize.py uses subprocess.run to invoke ffmpeg for audio decoding.
- [DYNAMIC_EXECUTION]: The skill uses dynamic imports in gen-video.example.mjs and gen-illustrations.mjs to load user-defined script and scene configurations (e.g., chapters.mjs, illustrations.mjs). This is a core part of its data-driven video generation pipeline.
- [EXTERNAL_DOWNLOADS]: The skill makes network requests to well-known service endpoints including generativelanguage.googleapis.com and api.elevenlabs.io to generate images and audio narration based on user scripts. These are established services and necessary for the primary functionality.
- [INDIRECT_PROMPT_INJECTION]: The skill processes potentially untrusted scripts (chapters.mjs) and recording marks (marks.txt) to automate video assembly. While it provides an esc() function for HTML sanitization in edit-engine.mjs, the ingestion of these inputs drives high-capability automation tasks like command execution and API interactions, representing a potential vulnerability surface.
Audit Metadata