The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill downloads the Whisper models and Python dependencies, including torch from the official PyTorch CPU wheel index at download.pytorch.org. These are expected resources for the skill's stated purpose.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection because it ingests untrusted audio data and converts it to text which is then added to the agent's context.
Ingestion points: User-provided audio files (e.g., audio.wav) processed by the local-whisper script.
Boundary markers: No explicit boundary markers or instructions to ignore embedded commands are present in the transcription workflow.
Capability inventory: The skill invokes a local script for transcription; however, the resulting text could influence downstream agent actions if the agent processes the transcript as instructions.
Sanitization: No sanitization or filtering of the transcribed text is described in the skill documentation.

local-whisper