PDF to Agent Skill

Pipeline

PDF → Split → Extract → Clean → Structure → Consolidate → Agent

Verktyg

PDF Split

pdftk input.pdf burst output page_%03d.pdf

Text Extraction

pdftotext -layout page.pdf page.txt

Python alternativ

# Split
from PyPDF2 import PdfReader, PdfWriter

# Extract (med tabeller)
import pdfplumber
with pdfplumber.open("page.pdf") as pdf:
    text = pdf.pages[0].extract_text()
    tables = pdf.pages[0].extract_tables()

Markdown-regler

Ta bort: Headers, footers, sidnummer, disclaimers

Behåll: Substans, definitioner, krav, siffror, tabeller

Strukturera:

# Rubrik
## Sektion
- Punktlista
| Tabell | Data |
**Term**: Definition

Konsolideringsmönster

Agent YAML

---
name: namn
description: beskrivning
tools: Read, Grep, Glob, Write, Edit
model: inherit
skills: namn
---

Skill YAML

---
name: namn
description: beskrivning
---

# Titel
[Snabbreferens - max 2-3 skärmar]

Filstruktur

output/
├── .claude/
│   ├── agents/[namn].md
│   └── skills/[namn]/SKILL.md
└── markdown/
    ├── [DOC]_COMPLETE.md
    └── [doc]_by_section/

Felsökning