pdf-extractlisted

# PDF Content Extraction Skill Extract PDF content to clean, organized markdown. ## Workflow 1. **Extract** — Run script to get raw content + metadata 2. **Analyse** — Review for patterns and issues 3. **Clean** — **Manually** remove noise (footers, watermarks, branding) 4. **Organise** — Restructure fragmented content 5. **Output** — Deliver clean markdown > **Note:** Only Step 1 uses a script. Steps 2–5 are performed manually by Claude reading and rewriting content. Do not write cleanup scripts. ## Step 1: Extract ```bash python /mnt/skills/user/pdf-extract/scripts/extract_pdf.py \ /mnt/user-data/uploads/{filename}.pdf \ /home/claude/extracted/ ``` **Options:** | Option | Description | |--------|-------------| | `--pages 1-10` | Extract specific page range | | `--method pymupdf4llm` | Force primary extractor (better formatting) | | `--method pymupdf` | Force fallback (more reliable for scanned PDFs) | | `--min-image-size 100` | Skip images smaller than 100px (filters icons) | **Output:** ``` /home/claude/extracted/ ├── {filename}.md # Raw markdown with YAML frontmatter ├── metadata.json # Structured metadata └── images/ # Extracted images (if any) ``` ## Step 2: Analyse Read the extracted markdown: ```bash cat /home/claude/extracted/{filename}.md ``` **Check YAML frontmatter for:** - `extraction_method` — Which extractor was used - `total_pages` — Document length - `has_outline` — Bookmarks exist (helps with structure) - `total_images