Voice Memo Transcribe

Transcribe Apple Voice Memos recordings to text, organize content, and save to Apple Notes. Workflow: iPhone Voice Memos → iCloud sync → Mac Voice Memos DB →...

This skill appears to do what it says: read your Mac's Voice Memos DB, transcribe audio locally with faster-whisper, and create Notes via AppleScript. Before installing: 1) Understand Full Disk Access grants the terminal broad file-read rights—only grant it if you trust the code and source. 2) faster-whisper may download large model files from the internet; expect network activity and disk usage. 3) Review the three included Python scripts yourself (they are small and readable) or run them manually on a test memo before automating. 4) If you prefer to limit risk, run the scripts in an isolated environment (temporary user account, VM, or container) and verify created Notes content. 5) Backup any important Notes/recordings before bulk operations.

Type: OpenClaw Skill Name: voice-memo-transcribe Version: 1.1.0 The skill accesses sensitive user data by reading the Apple Voice Memos SQLite database (~/Library/Group Containers/.../CloudRecordings.db) and audio files, which requires Full Disk Access. While the Python scripts (list_memos.py, transcribe.py, extract_transcript.py) appear to be legitimate tools for transcription, the instructions in SKILL.md for saving notes are highly vulnerable to injection. Specifically, the use of unquoted shell heredocs (<<< "$NOTE_BODY") and direct string interpolation in AppleScript (set noteTitle to "<TITLE>") allows content from the transcription to potentially trigger shell or AppleScript execution if the AI agent does not perform rigorous sanitization.