The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Interacts with ElevenLabs' official API endpoints for cloud-based voice synthesis and transcription tasks.
[EXTERNAL_DOWNLOADS]: Fetches pre-trained machine learning models from Hugging Face for local transcription using the mlx-whisper library.
[COMMAND_EXECUTION]: Invokes macOS system binaries say and afplay via safe subprocess calls to facilitate local text-to-speech and audio playback.
[PROMPT_INJECTION]: The skill presents an indirect prompt injection surface where audio data is ingested and transcribed into the agent's context.
Ingestion points: Audio transcription results from scripts/stt_elevenlabs.py and scripts/stt_local.py.
Boundary markers: Absent; the transcripts are output as raw strings without delimiters or instructions to ignore embedded commands.
Capability inventory: Includes the ability to write files, make network requests to ElevenLabs, and execute system commands for audio feedback.
Sanitization: Transcribed content is not filtered or sanitized before being returned to the agent.

voice