The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of the whisper command-line interface for transcribing and translating audio files locally on the user's system.- [EXTERNAL_DOWNLOADS]: The skill automates the installation of the openai-whisper tool using the Homebrew package manager. Additionally, the CLI tool is designed to download necessary transcription models from OpenAI's servers to the ~/.cache/whisper directory upon its first execution.- [INDIRECT_PROMPT_INJECTION]: The skill processes external audio data which, when transcribed into text, is introduced into the agent's context. This creates a potential vulnerability where malicious instructions embedded in audio could influence the agent's behavior.
Ingestion points: Audio files provided by the user (e.g., .mp3, .m4a) are read and transcribed via the whisper CLI.
Boundary markers: There are no explicit delimiters or instructions provided to the agent to disregard commands found within the transcribed text.
Capability inventory: The skill has the capability to execute shell commands (whisper).
Sanitization: No filtering or validation is performed on the output of the transcription before it is returned to the agent.

openai-whisper