The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes ffmpeg via the Python subprocess module in scripts/transcribe_groq.py for audio compression and in scripts/resegment.py for silence detection. These calls utilize list-based arguments, which is a security best practice that prevents shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The transcription script scripts/transcribe_groq.py makes network requests to api.groq.com. This is a well-known service provider and the communication is essential for the skill's advertised transcription functionality.
[CREDENTIALS_UNSAFE]: The skill manages the Groq API key securely by attempting to load it from the GROQ_API_KEY environment variable or a specific file in the user's home directory (~/.groq_api_key), rather than using hardcoded secrets.
[SAFE]: The skill's architecture is focused and transparent. It includes specific scripts for processing, cleaning, and validating subtitle data, all of which operate locally on user-provided media files and transcription results without exhibiting malicious patterns like obfuscation or persistence.

audio-to-srt