Convert audio, Office documents, and scanned files into Markdown for wiki ingestion.
npx skills add https://github.com/cookkie03/skills --skill wiki-preprocess-cookkie03
Please help me install this Agent Skill. Skill: wiki-preprocess Source: https://github.com/cookkie03/skills/tree/main/wiki-preprocess Command: npx skills add https://github.com/cookkie03/skills --skill wiki-preprocess-cookkie03
<name>.transcription.md next to the original. - Office document conversion: Uses markitdown to convert docx, pptx, xlsx, odt, epub, csv, html, and other structured formats into <filename>.md files preserving headings, tables, and lists. - OCR delegation: Routes scanned images and PDFs to a dedicated OCR skill, producing <name>.ocr.md outputs. - Use Case: You drop a recorded voice memo and a quarterly report docx into your vault's raw/ folder; run the preprocessing scripts to generate transcription and Markdown files that wiki-ingest can read directly. ## Quick Start Ask the assistant to preprocess the audio and Office documents in your vault's raw folder so they are ready for wiki ingestion.High-intent search queries and answers about installing and using this skill.
Run the preprocess-audio.py script, which converts audio files to MP3 with ffmpeg and transcribes them using Whisper. It saves a <name>.transcription.md file next to each audio file and skips files that are already up to date.
Use the preprocess-office.py script, which converts Office and structured documents with the markitdown library. It preserves headings, tables, and lists, writing the result as <filename>.md next to the original document.
Legacy binary Office formats often fail with markitdown directly. Convert them first with LibreOffice using libreoffice --headless --convert-to docx, then rerun the script, which flags these cases automatically.
The script handles m4a, opus, wav, ogg, flac, aac, wma, and mp3 files. Non-MP3 files are first converted to 16kHz mono MP3 with ffmpeg before transcription with the Whisper turbo model.
Scanned PDFs contain images rather than text layers, so markitdown cannot extract content. Route them through an OCR skill to generate a <name>.ocr.md file; native-text PDFs can be read directly without OCR.
ssdavidaiTransform inbound emails, transcripts, and documents into structured vault records with YAML frontmatter.
moltis-orgIngest sources into an interlinked Markdown knowledge base with schema enforcement.
BEIRUXPerform LLM-optimized web search and content extraction with JSON output.
xingBaGanAnalyze video content and transcribe audio locally with ffmpeg and Whisper.