Transcribe Basic

Transcribe Basic

Metadata

• Trigger when: the user needs a fast transcript without speaker labels.
• Do not use when: the user explicitly needs diarization or known-speaker matching.

Skill Purpose

Generate the fastest honest transcript path for audio by keeping the job to plain transcription instead of overcomplicating it with diarization.

Instructions

1. Confirm the audio path and check that OPENAI_API_KEY is set. If the key is missing, stop and tell the user to set it locally.
2. Run ~/.codex/skills/transcribe/scripts/transcribe_diarize.py with gpt-4o-mini-transcribe and plain text output under a stable output path such as output/transcribe/.
3. Validate that the transcript exists, is readable, and matches the requested file/output format. Note any caveat caused by audio quality, chunking, or language uncertainty.

Non-Negotiable Acceptance Criteria

• Secrets stay in environment variables, never in chat.
• The model choice is the fast plain-transcription path, not diarization-heavy overkill.
• Output location and format are explicit.
• Audio-quality limitations are stated when they materially affect trust in the transcript.

Output

• The transcript path.
• The model and response format used.
• A short note on transcript quality or caveats.
• Next skill options (only if needed): $transcribe-diarize — use when the user later needs speaker labels or diarization.