The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a local Python script (transcribe.py) to process media files. The script uses the argparse library to handle inputs and does not use dangerous shell execution functions (like os.system or subprocess.run with shell=True) on untrusted data. Input paths are validated and resolved using the pathlib library.
[EXTERNAL_DOWNLOADS]: The skill utilizes the openai-whisper library, which downloads pre-trained models from OpenAI's official repositories upon first use. These are trusted sources and necessary for the skill's primary function.
[DATA_EXFILTRATION]: No evidence of unauthorized data transmission was found. The transcription process occurs entirely on the local machine using the local Whisper installation. Output files are saved to the project directory or a location specified in the .env file.
[PROMPT_INJECTION]: The SKILL.md file contains instructions for the AI on how to summarize the resulting transcript. While it includes 'tone guidelines' to avoid negative phrasing, these are used for stylistic consistency and do not attempt to bypass core AI safety filters or override system instructions.

media-transcriber