AgentSkillsCN

CapturePdf

将 PDF 书籍捕获并导入 AI 大脑。当用户希望捕获 PDF、导入 PDF、添加 PDF 书籍，或对 PDF 进行嵌入式处理时，可使用此功能——输入 /capture-pdf 或 /pdf。

SKILL.md
--- frontmatter
name: CapturePdf
description: Capture PDF books to ai-brain sources. USE WHEN user wants to capture pdf, import pdf, add pdf book, process pdf for embeddings OR /capture-pdf OR /pdf.

CapturePdf

Captures PDF books and converts them to markdown for the ai-brain knowledge base.

How It Works

  1. •Extracts text from all PDF pages
  2. •Cleans content (removes tracking, formatting artifacts)
  3. •Saves as markdown in sources/
  4. •Auto-commits to git
  5. •Ready for embedding via embed_sources.py

Usage

Basic (interactive - will prompt for author/title if not in PDF metadata):

bash
python3 ~/ai-brain/scripts/capture_pdf.py /path/to/book.pdf

With metadata (recommended for cleaner filenames):

bash
python3 ~/ai-brain/scripts/capture_pdf.py /path/to/book.pdf --author "Author Name" --title "Book Title"

Without auto-commit:

bash
python3 ~/ai-brain/scripts/capture_pdf.py /path/to/book.pdf --no-commit

After Capture

Generate embeddings for the new content:

bash
python3 ~/ai-brain/scripts/embed_sources.py

Examples

Example 1: Capture a fitness book

code
User: "/capture-pdf"
User provides: /home/user/downloads/531-forever.pdf
→ Runs capture script with --author "Jim Wendler" --title "5/3/1 Forever"
→ Creates sources/2026-01-18-jim-wendler-531-forever.md
→ Commits to git
→ User runs embed_sources.py to index

Example 2: Capture with prompts

code
User: "capture this pdf ~/books/some-book.pdf"
→ Runs capture script
→ Script prompts for author/title if not detected
→ Creates markdown in sources/
→ Ready for embeddings

Example 3: Batch capture

code
User: "I have 3 Wendler PDFs to capture"
→ Run capture script for each PDF
→ Then run embed_sources.py once at the end