Lecteur et analyseur PDF

VérifiéPrudence

Lire, extraire et analyser le contenu de documents PDF. Résumer, extraire des informations spécifiques, comparer plusieurs PDF et convertir les données en formats structurés.

Spar Skills Guide Bot
ProductiviteDébutant
2002/06/2026
Claude Code
#pdf#document-analysis#text-extraction#summarization

Recommandé pour

Notre avis

Cette compétence permet de lire, extraire et analyser le contenu de documents PDF via des commandes simples.

Points forts

  • Lecture native des PDF sans outil externe
  • Extraction ciblée d'informations (tableaux, citations, contacts)
  • Capacité de résumé et de comparaison entre plusieurs PDF

Limites

  • Les PDF scannés sans OCR ne peuvent pas être extraits correctement
  • Certains PDF protégés empêchent l'extraction de texte
  • Les très gros PDF (100+ pages) peuvent nécessiter une pagination manuelle
Quand l'utiliser

Utilisez cette compétence lorsque vous devez extraire ou résumer le contenu textuel d'un fichier PDF.

Quand l'éviter

Évitez de l'utiliser pour des PDF contenant uniquement des images scannées sans texte sous-jacent, car l'OCR n'est pas intégré.

Analyse de sécurité

Prudence
Score qualité90/100

The skill primarily uses the built-in Read tool for PDF files, which is low-risk. However, it also instructs users to run a Python utility via `uv run`, a command that executes a Python script. While the script likely performs benign PDF operations, any command-line execution carries potential risk if the tool is maliciously modified or input is not sanitized. No destructive or exfiltration commands are directly present.

Points d'attention
  • Uses a bash command `uv run tools/pdf_util.py` which invokes a Python script; if the script or its dependencies are compromised, it could lead to arbitrary code execution. However, the script is local and intended for PDF processing, so risk is moderate.

Exemples

Summarize a PDF
Read /path/to/document.pdf and give me a concise summary of the main points.
Extract tables
Extract all tables from /path/to/report.pdf and display them as markdown.
Compare two PDFs
Compare /path/to/draft1.pdf and /path/to/draft2.pdf and list the differences.

name: pdf description: This skill should be used when the user asks to "read pdf", "view pdf", "extract text from pdf", "summarize pdf", or shares a PDF file path.

/pdf

Read, extract, and analyze PDF documents.

Instructions

When the user provides a PDF path or asks about PDF content:

1. Reading PDFs

Use the Read tool directly on the PDF file. Claude Code supports reading PDF files natively:

Read: /path/to/document.pdf

This extracts both text and visual content for analysis.

2. Common Tasks

Summarize: Provide a concise summary of the document

  • Key points and main arguments
  • Document structure overview
  • Important figures, tables, or data

Extract specific info: Find and extract:

  • Tables (convert to markdown format)
  • Code snippets
  • Citations/references
  • Contact information
  • Dates and deadlines

Compare PDFs: When given multiple PDFs:

  • Identify differences
  • Highlight common themes
  • Cross-reference information

3. Python Utility

For batch operations or programmatic access:

uv run tools/pdf_util.py extract /path/to/file.pdf
uv run tools/pdf_util.py info /path/to/file.pdf
uv run tools/pdf_util.py search /path/to/file.pdf "search term"

4. Output Format

When presenting PDF content:

  • Use markdown headers for document sections
  • Convert tables to markdown tables
  • Note page numbers for reference: (p. 5)
  • Flag any extraction issues (scanned images, encrypted content)

5. Limitations

  • Scanned PDFs may have limited text extraction (OCR not included)
  • Some PDFs have copy-protection that prevents text extraction
  • Very large PDFs (100+ pages) may need pagination
Skills similaires