The Agent Skills Directory

[COMMAND_EXECUTION]: The skill implements its core functionality via scripts/check.py and scripts/fix_sound.py. These scripts execute external binaries (ffmpeg and ffprobe) using the Python subprocess module.
The execution calls use list-based arguments rather than shell strings, which effectively prevents shell injection vulnerabilities.
Command parameters are derived from specific video measurements (floats/integers) or well-defined user inputs.
[EXTERNAL_DOWNLOADS]: The skill includes evaluation scripts (place-files.sh) that download sample video files for testing purposes.
These downloads target edit.trysometh.ing, which is a vendor-owned domain associated with the author editatlas.
The skill instructions explicitly direct the agent to work with local files and advise against downloading from untrusted social media platforms.
[INDIRECT_PROMPT_INJECTION]: The skill possesses an attack surface where it processes external video file metadata and streams.
Ingestion points include file probing and audio analysis outputs from FFmpeg.
Capability inventory includes the ability to generate reports and modify audio streams using ffmpeg.
The skill mitigates risks by parsing tool outputs using regex and structured JSON loading, and it uses specific boundary templates (REVIEW.md) to report findings to the user.

video-check