The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill possesses an ingestion surface for untrusted data through audio transcription (STT) and text narration (TTS).
Ingestion points: The skill ingests local audio files (mp3, wav, m4a, webm, flac) via the voicebox_transcribe MCP tool and text input via voicebox_speak.
Boundary markers: The instructions do not define specific delimiters or "ignore embedded instructions" warnings for the processed transcription output before it is returned to the agent context.
Capability inventory: The skill is capable of writing audio files and manifests to the local filesystem and executing local command-line utilities.
Sanitization: While the skill recommends stripping markdown and emojis for audio quality, it does not describe security-focused sanitization or filtering of transcribed text.
[COMMAND_EXECUTION]: The skill intends to execute the local ffmpeg utility to perform audio transcoding (converting WAV to MP3) for generated assets.
[EXTERNAL_DOWNLOADS]: The documentation directs users to download and install the external Voicebox desktop application from its public GitHub repository (github.com/jamiepine/voicebox/releases).

oma-voice