The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides templates for executing PowerShell commands to run whisper, ffprobe, and ffmpeg locally. These commands utilize environment variables (e.g., $env:PYTHONIOENCODING='utf-8') and pass file paths as arguments. While intended for local transcription, the use of shell commands creates a vector for command injection if file paths or arguments are not properly sanitized by the agent.
[INDIRECT_PROMPT_INJECTION]: The skill processes external media files (MP4, MP3, etc.) to generate text transcripts. These transcripts are then used as context for the agent to generate meeting minutes, action items, and summaries. This represents an indirect prompt injection surface where malicious content embedded in the audio/video (via speech-to-text) could attempt to influence the agent's summary or behavior.
Ingestion points: Local media files (MP4, WAV, etc.) read via whisper CLI (SKILL.md, references/transcription-workflow.md).
Boundary markers: None explicitly defined in the instructions for separating transcript content from agent instructions during the summary phase.
Capability inventory: The skill uses subprocess calls to whisper, ffmpeg, and ffprobe. It also performs file reads and writes (SKILL.md).
Sanitization: No specific sanitization or filtering of the generated transcript text is mentioned before it is processed by the LLM for summarization.

local-media-transcription