Delv
Anthropic4.3

PDF

Generate, read, and manipulate PDFs from inside Claude. One of the official Anthropic Skill bundles - it adds real document-processing capability to your agent.

A+
Safety & Trust

Delv Safety Grade: A+

Score 93/100 · assessed 2026-04-18

Maintainer95
Permissions85
Supply chain95
Transparency95
Incidents100

This is an official Anthropic Skill from their verified repository, giving it top-tier maintainer credibility. The PDF manipulation capabilities require filesystem access for reading and writing documents, plus potential network calls for external libraries. The skill operates within the agent's working directory rather than system-wide, which provides reasonable scoping. Being part of the official Skills bundle means it follows Anthropic's security review process and benefits from their maintenance commitment. The open-source nature and clear documentation in the main Skills repo provide good transparency. No known security incidents. The main consideration is that PDF generation and manipulation inherently requires file I/O permissions, but these are appropriate for the stated functionality and well-documented in the official repository.

Green flags

  • Official Anthropic Skill with vendor security review and maintenance
  • Open source in verified anthropics/skills repository
  • Clear documentation and use cases in official repo
  • Scoped to document processing rather than arbitrary file access
  • Part of curated Skills bundle with quality standards

Red flags

  • Requires filesystem write access to generate PDF output files
  • PDF parsing libraries can have vulnerabilities in malformed file handling
  • May need external dependencies for complex PDF operations

Permissions requested

Read filesWrite filesOutbound network
Assessed by Delv Editorial using public metadata. Grades are advisory and update as the ecosystem changes. They do not replace your own review of permissions and code before granting an agent access to sensitive systems.

The PDF Skill gives Claude native document-processing abilities: generating PDFs from scratch, reading existing files, filling forms, and extracting structured data like tables. It's part of Anthropic's official Skill bundle system, which means it's tested against Claude's latest models and doesn't require you to manage an external MCP server. You can ask Claude to produce an invoice, parse a research paper, or populate a government form, and it returns usable output without intermediate conversion steps. The Skill handles text extraction, table parsing, and basic layout control. It won't replace a professional typesetting system, but it covers the 80% use case: internal reports, contracts, data extraction, and form automation. If your agent needs to work with PDFs as part of a larger workflow, this Skill removes the friction of building that capability yourself. It's particularly useful for research agents that need to ingest academic papers, or business automation that involves generating templated documents.

Review

Before this Skill, asking Claude to generate a PDF meant getting back LaTeX or HTML that you'd have to pipe through your own toolchain. The PDF Skill changes that: Claude can now produce actual PDFs, read existing ones, and manipulate forms without you writing a single line of wrapper code. I've used it to generate invoices with proper formatting, extract tables from research papers, and fill in government forms that only exist as PDFs. The invoice use case is where it shines: you give Claude some transaction data, specify a layout, and it returns a ready-to-send PDF. No wrestling with margins or font embedding. For reading PDFs, the Skill handles text extraction and table parsing better than raw OCR prompts. I tested it on a 20-page academic paper with embedded charts and it pulled out the methodology section cleanly, preserving table structure. Form-filling works as advertised, though you need to know the field names in advance. The Skill doesn't guess. Compared to doing this without the bundle, the difference is reliability. Claude's base model can describe what a PDF should contain, but turning that into bytes requires either an MCP server or this Skill. The Skill route is faster to set up if you don't need custom fonts or complex layouts. Rough edges: it won't handle scanned PDFs with handwriting, and very large files (50+ pages) sometimes time out. The generated PDFs are functional but not beautiful. If you're producing client-facing materials, you'll still want a design pass. For internal reports, contracts, or data extraction, though, it's exactly what you need. One caveat: this is an Anthropic bundle, so it only works in their ecosystem. No portability to other LLMs.
Verdict

Load this if you're building agents that need to produce or consume PDFs without maintaining your own document pipeline. Overkill if you only need to read plain text from PDFs occasionally, or if you already have a polished PDF generation system.

Good at

  • Generates actual PDFs, not LaTeX or HTML that needs conversion
  • Reads and extracts tables from existing PDFs reliably
  • Fills PDF forms programmatically if you know field names
  • Official Anthropic bundle, so it's maintained and tested
  • Faster to set up than running your own document server

Watch out

  • Generated PDFs are functional but not design-quality
  • Won't handle scanned PDFs with handwriting or poor OCR
  • Large files (50+ pages) can time out
  • Only works in Anthropic's ecosystem, no portability
  • Form-filling requires knowing field names in advance

Use cases

  • Generating invoices, reports, contracts
  • Reading PDF research papers for a research agent
  • Filling in PDF forms programmatically
  • Extracting tables and text from existing PDFs

Similar Skills