菜单
Transcribe audio files to text. Use when you need to convert speech/audio to text.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
Use this skill for any PDF authoring or post-processing task that doesn't require READING the content. CREATE / GENERATE / PRODUCE PDFs from scratch — reports, invoices, business letters, memos, status updates, market analyses, proposals (Typst engine, charts via Cetz, four bundled templates: report, invoice, letter, memo). FILL existing PDF forms — text fields, checkboxes, radio buttons. MERGE / SPLIT / ROTATE / WATERMARK / ENCRYPT existing PDFs via qpdf and pypdf. Embed ZUGFeRD / Factur-X XML for EU e-invoices. Triggers: 'erstelle PDF', 'baue Rechnung', 'mache einen Bericht als PDF', 'create a PDF', 'generate a report', 'render an invoice', 'business letter', 'Geschäftsbrief', 'Memo', 'merge PDFs', 'split a PDF', 'rotate pages', 'watermark this PDF', 'encrypt PDF', 'PDF Formular ausfüllen', 'fill this form'. Do NOT use for READING / EXTRACTING text from existing PDFs or OCR on scanned documents — use the `document-parse` skill (LiteParse).
Schedule events that fire at a specific time, when an email reply arrives, or when a shell check passes. Use to wake a session without idle polling.
Use this skill whenever the user wants to create, read, edit, or manipulate Word documents (.docx files). Triggers include: any mention of 'Word doc', 'word document', '.docx', or requests to produce professional documents with formatting like tables of contents, headings, page numbers, or letterheads. Also use when extracting or reorganizing content from .docx files, inserting or replacing images in documents, performing find-and-replace in Word files, working with tracked changes or comments, or converting content into a polished Word document. If the user asks for a 'report', 'memo', 'letter', 'template', or similar deliverable as a Word or .docx file, use this skill. Do NOT use for PDFs, spreadsheets, Google Docs, or general coding tasks unrelated to document generation.
How to create and manage triggers via the CLI. Covers cron, webhook, manual, Signal, WhatsApp and Email integration.
Use this skill to READ / EXTRACT / PARSE / OCR existing documents — pull text and structure out of PDF, DOCX, PPTX, XLSX, EPUB, HTML, images (JPG/PNG/TIFF/WebP), scanned-document PDFs, or any unstructured file. Triggers: 'parse this PDF', 'extract text', 'OCR this image', 'text aus scan', 'lies das Dokument', 'convert PDF to markdown', 'erkenne Text', 'transcribe document', 'spatial layout extraction', plus any time the user provides a file and asks what's inside. Runs LiteParse locally — Tesseract.js OCR built-in, optional external OCR server (EasyOCR / PaddleOCR) for higher accuracy. Output is clean markdown ready for downstream LLM processing. Do NOT use for: (1) CREATING a new PDF from scratch — use the `pdf` skill. (2) Filling forms in an existing PDF — see `pdf` skill `forms.md`. (3) Pure speech-to-text from audio/video — use the `stt` skill.
Use this skill to CUT, TRIM, CLEAN, EDIT existing video footage. Triggers: 'schneide das Video', 'edit/cut/trim this video', 'remove silence/fillers', 'concat videos', 'add subtitles', 'edit these into a launch video', 'make a recap from these clips', any task that takes existing video file(s) as input and produces a polished edited output. Audio-first approach: transcribe → identify cuts on word boundaries → FFmpeg-based EDL render → self-evaluate cut points. Do NOT use for: (1) creating animated videos from scratch — use the `video` skill (Remotion). (2) understanding/classifying video content (describe scenes, extract timestamps for visual events, classify) — use a multimodal LLM directly. (3) raw single-command FFmpeg ops where you already know the exact filter chain — just run FFmpeg directly.
| name | stt |
| description | Transcribe audio files to text. Use when you need to convert speech/audio to text. |
Transcribe audio files to text using the stt CLI.
# Transcribe an audio file
stt /path/to/audio.wav
# Transcribe with language hint
stt --language de /path/to/audio.ogg
# Transcribe from a URL (downloads first)
stt https://example.com/recording.mp3
wav, mp3, ogg, m4a, aac, flac, webm, mp4 (audio track)
The STT endpoint is Whisper-compatible (OpenAI /v1/audio/transcriptions format):
curl -X POST "$STT_URL" \
-F "file=@/path/to/audio.wav" \
-F "response_format=json" \
-F "language=de"
Response: {"text": "transcribed text here"}
The URL is resolved from: ATLAS_STT_URL env → STT_URL env → config.yml (stt.url) → default.
Audio messages received via Signal are automatically transcribed before reaching the agent — no manual action needed.