The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external audio files and returns the resulting transcription directly to the agent's context. This creates a surface for indirect prompt injection where spoken instructions within the audio could influence the agent's behavior.
Ingestion points: The AUDIO_FILE argument passed to transcribe.sh is the source of untrusted data.
Boundary markers: No delimiters or safety instructions are used when returning the transcription text to the agent.
Capability inventory: The skill uses curl for network requests and reads local files.
Sanitization: The output from the Groq API is returned as raw text without sanitization or filtering.
[PERSISTENCE]: The setup instructions in SKILL.md guide the user to modify their shell configuration files (~/.zshrc or ~/.bashrc) to store the GROQ_API_KEY. While common for developer tools, this modifies the user's environment across sessions.
[EXTERNAL_DOWNLOADS]: The transcribe.sh script uses curl to send audio data to https://api.groq.com/openai/v1/audio/transcriptions. Groq is a well-known AI infrastructure service, and this operation is consistent with the skill's stated purpose.

transcribe