ow-tools

原文涵盖的主题

• /ow-tools
• Steps
• What ow-tools does
• What it never does

ow-tools runs code on this machine. The ow- skills are text only; every piece of code they need is here. Without ow-tools they still work, using a transcript, notes or Markdown instead. 1. Run ow-tools doctor --quick and show its output as it is. It checks each requirement and prints the command that fixes it. Run plain ow-tools doctor only when the user asks to verify the Whisper model's checksum (it reads 1.6 GB). 2. Explain each MISS line in one plain sentence. Group them by what they unlock: transcription, PDF and Word export, PowerPoint export, screenshots (capture), or the secrets scan (gitleaks and git). An opt line (pip-audit) is optional. 3. The user runs the fixes in their own terminal. Never run brew, apt, sudo, uv sync, pipx, a model or browser download or a keychain command yourself. The commands must match the ones in the ow-tools README. The Hugging Face token is needed once, for the speaker-model fetch. The user stores it in their keychain. It never goes in a file, a repository or this chat. 4. After the user has run the fixes, run ow-tools doctor again. | Command | For | |---|---| | ow-tools transcribe /audio/[ ] | A recording → transcript.json + transcript.md, loc

这里只展示节选，请以完整原文和当前要求为准。