メニュー
Transcribe macOS meeting audio from the project folder with a bundled whisper.cpp CLI, downloading the selected model on first use, and write transcript files back to the project folder.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Unlocks pdf_author and pdf_author_edit to create and revise professional print-ready PDF reports, status packs, risk reports, and fixed-layout client deliverables from the TensorPM project. A planner structures only sourced facts, a print design director defines the editorial system, an HTML author composes the report, Chromium renders it, and an independent structural/visual reviewer drives automatic repair. Also guides targeted PDF stamping, merging, splitting, and edits.
Unlocks document_author and document_author_edit to create and revise professional editable Word/DOCX reports, meeting minutes, decision memos, requirements documents, and risk reports from the TensorPM project. A planner structures only sourced project facts, a document design director defines the editorial system, then an author composes the DOCX and an independent reviewer drives automatic repair.
Use this skill when reviewing, auditing, visually checking, scoring, or improving generated TensorPM artifacts and deliverables, including PowerPoint decks, Excel workbooks, Word documents, PDF reports, previews, Folien, Tabellen, Berichte, Protokolle, Druckversionen, structured diagnostics, render-and-critique loops, artifact QA, quality gates, and acceptance review.
Unlocks the deck_author and deck_author_edit tools — generate and edit any kind of PowerPoint presentation (status report, pitch, overview, proposal, review, kickoff) from the TensorPM project. A planner builds the narrative from the project graph (invents nothing), an art director commits to a design, then the deck is freely composed, sandbox-rendered and auto-repaired.
Use this skill any time Excel, XLSX, spreadsheet, Tabelle, Arbeitsmappe, workbook, budget tracker, milestone schedule, risk register, action-item export, formulas, editable calculation model, reconciliation, or spreadsheet QA is involved.
Parse German GAEB bills of quantities (Leistungsverzeichnis, .x81/.x83/.x84) deterministically and stage their positions for review as project material items with position references, quantities, and unit prices.
| name | meeting-transcriber |
| description | Transcribe macOS meeting audio from the project folder with a bundled whisper.cpp CLI, downloading the selected model on first use, and write transcript files back to the project folder. |
| version | 0.1.1 |
| runtime | {"engine":"deno"} |
| permissions | {"tensorpm":{"action_items":"none","decisions":"none","history":"none"},"network":["huggingface.co:443","cdn-lfs.hf.co:443","cas-bridge.xethub.hf.co:443","transfer.xethub.hf.co:443"],"project_files":{"read":["meetings/**","skill-cache/meeting-transcriber/**","transcripts/**"],"write":["transcripts/**","skill-cache/meeting-transcriber/**"]},"run":["system:afconvert","skill:assets/bin/darwin-arm64/whisper-cli"],"ai_provider":"none"} |
| limits | {"cpu_seconds":300,"memory_mb":2048} |
| scripts | {"transcribe":{"entrypoint":"scripts/transcribe.ts","description":"Decode a project audio file with macOS afconvert, run the bundled whisper.cpp CLI, and save transcript artifacts under transcripts/.","schema":{"input":{"type":"object","properties":{"audio_path":{"type":"string","description":"Project-relative path under meetings/, e.g. meetings/client-sync.m4a."},"model":{"type":"string","enum":["tiny","base","small"],"description":"Whisper model to download/use from the local project cache. Defaults to base."},"language":{"type":"string","description":"Optional Whisper language code, e.g. de or en."}},"required":["audio_path"]},"output":{"type":"object","properties":{"audio_path":{"type":"string"},"model":{"type":"string"},"transcript_markdown_path":{"type":"string"},"transcript_text_path":{"type":"string"},"transcript_vtt_path":{"type":"string"},"transcript_json_path":{"type":"string"}},"required":["audio_path","model","transcript_markdown_path","transcript_text_path"]}}}} |
Use this skill when a user has placed a meeting audio file under meetings/
in the project folder and wants a local transcript written under
transcripts/.
This is intentionally macOS-only for the first version:
afconvert decodes the source audio to 16 kHz mono WAV.whisper.cpp whisper-cli binary must be shipped in the installed skill
payload at assets/bin/darwin-arm64/whisper-cli.skill-cache/meeting-transcriber/models/.The script returns project-relative paths for the generated Markdown, TXT, VTT, and JSON outputs. It does not create TensorPM action items by itself; the agent should read the transcript and use TensorPM's existing proposal or action-item tools for human-reviewed meeting notes.