The Agent Skills Directory

[COMMAND_EXECUTION]: The srt_helper.py script utilizes subprocess.run to invoke ffmpeg for audio processing and pause insertion. This is implemented using a list-based argument structure, which is the recommended safe practice to prevent shell injection vulnerabilities.
[EXTERNAL_DOWNLOADS]: The shell scripts listenhub-tts.sh and listenhub-speakers.sh download audio files, subtitle data, and voice lists from the ListenHub API (api.marswave.ai). The skill also connects to the Groq and OpenAI APIs for transcription services. These external references are appropriate for the skill's purpose and target established service endpoints.
[INDIRECT_PROMPT_INJECTION]: The skill provides a subtitle correction mechanism that interpolates user-supplied text (transcribed captions) into a prompt for an LLM. While this is an indirect injection surface, it is mitigated by explicit system instructions that constrain the LLM to only return a specific JSON format and strictly correct text without following embedded commands.

verysmallwoods-video