The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/render_edl.py uses subprocess.run to call ffmpeg and ffprobe for media processing.
Evidence: Multiple calls in the run(), probe_duration(), trim_clip(), build_audio_bed(), and render() functions.
Security Note: Command arguments are passed as lists rather than shell strings, which significantly mitigates shell injection vulnerabilities. Additionally, numerical inputs from the EDL file are cast to float or int before being used in command construction.
[DATA_EXFILTRATION]: No network operations or data exfiltration patterns were identified. The skill operates entirely on local files.
[CREDENTIALS_UNSAFE]: No hardcoded API keys, tokens, or other sensitive credentials were found in the source code or documentation.
[PROMPT_INJECTION]: The skill instructions and metadata do not contain attempts to override agent behavior or bypass safety guidelines.

render-edl