The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides shell commands for installing necessary packages and interacting with the Whisper CLI.
Evidence: Shell blocks in SKILL.md include pip install -U openai-whisper, brew install ffmpeg, sudo apt install ffmpeg, and various whisper command-line examples.
[PRIVILEGE_ESCALATION]: The installation guide for Ubuntu systems involves the use of sudo to acquire administrative privileges for package installation.
Evidence: The command sudo apt install ffmpeg is listed in the Installation section of SKILL.md to fulfill a system dependency.
[INDIRECT_PROMPT_INJECTION]: The skill processes external audio and video files for transcription, creating a surface where adversarial audio could potentially influence the agent's output or downstream tasks.
Ingestion points: model.transcribe("audio.mp3") and similar functions in SKILL.md and references/languages.md process user-provided media files.
Boundary markers: The provided code examples do not demonstrate the use of delimiters or specific instructions to the model to ignore embedded commands in the audio.
Capability inventory: The skill generates text transcriptions and translations that can be printed to the console or saved to the file system.
Sanitization: There is no evidence of text sanitization or validation performed on the transcription results before they are returned or stored.

whisper