sync-narration
Pass
Audited by Gen Agent Trust Hub on Oct 28, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSDYNAMIC_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill executes external command-line tools including
ffmpegandffprobeto perform video editing, frame extraction, and metadata inspection. These operations are performed usingspawnSyncwith argument arrays, which is a standard and secure method for preventing shell command injection. - [EXTERNAL_DOWNLOADS]: The skill's installation process involves downloading the
faster-whisperPython package. On its first execution, this library automatically fetches official pre-trained machine learning models from a public repository to facilitate local speech-to-text alignment. - [DYNAMIC_EXECUTION]: The synchronization script (
sync-narration.mjs) dynamically assembles a complex ffmpeg filter graph at runtime. This graph is constructed based on the timing data generated in previous steps to precisely align audio segments, insert silence gaps, and overlay styled captions. - [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided text files for script and caption content. It mitigates potential injection risks by writing the caption text to temporary files and referencing them using ffmpeg's
textfileparameter, ensuring the data is interpreted strictly as literal text for display rather than executable filter instructions. - [SAFE]: No obfuscation, data exfiltration, hardcoded credentials, or persistence mechanisms were found. The skill operates entirely on local media assets as described in its documentation.
Audit Metadata