CookiesForAI

pdf

PDF (OpenAI)

pdf is OpenAI's open-source agent skill for making and reviewing PDF files: the agent builds them with Python libraries, renders each page to an image and checks the layout before delivering.

Agent skill by OpenAI Official · Updated

Install PDF (OpenAI)

Install the skill with the skills CLI

Terminal
npx skills add openai/skills --skill pdf

Or, inside Codex, use the built-in skill installer

Terminal
$skill-installer pdf

Restart Codex after installing so it picks up the new skill.

Or paste this into your coding agent: Install the agent skill pdf from github.com/openai/skills. A skill can include scripts that run on your computer, so read its source first.

What PDF (OpenAI) does

The PDF skill from OpenAI's skills catalog tells an agent how to handle PDF files where layout matters. For new documents it uses the reportlab Python library; for reading text and running quick checks it uses pdfplumber or pypdf, while warning that text extraction says nothing reliable about how the page looks.

Its main rule is visual review. After every meaningful change, the agent renders pages to PNG images with Poppler's pdftoppm and inspects them for clipped text, overlapping elements, broken tables, black squares or unreadable characters. It is told not to deliver until the latest render shows no defects, and to check headers, footers and page numbers.

The skill also fixes housekeeping: intermediate files go in tmp/pdfs and final files in output/pdf, with stable descriptive names. If a dependency cannot be installed, the agent is told to say which one is missing and how to install it, rather than carry on without checking the output.

When to use PDF (OpenAI)

  • You want an agent to produce a report or invoice as a PDF that looks right on every page.
  • You need to check a generated PDF for clipped text or broken tables before sending it.
  • You want text pulled out of a PDF for a quick check.
  • You want an open-licensed PDF skill you can copy and adapt.

When to pick something else

  • Filling PDF forms, splitting or merging files, or OCR on scans: the skill does not cover them. Anthropic's PDF skill does, under a proprietary license.

What PDF (OpenAI) needs

  • Python with reportlab, pdfplumber and pypdf
  • Poppler (pdftoppm) for rendering pages to images

Which agents PDF (OpenAI) works in

OpenAI documents PDF (OpenAI) for Codex, Any agent that reads SKILL.md files, through the skills CLI. The open skills CLI also installs it into 78 agents, including Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode (we listed it with the CLI on October 1, 2026). See where each agent looks for skills.

OpenAI marked openai/skills deprecated on 2026-06-22 and points to its openai/plugins repository for current Codex skills. This copy still installs, but may not get updates.

PDF (OpenAI) license

PDF (OpenAI) is published under Apache-2.0. The skill folder carries its own Apache-2.0 LICENSE.txt, so you may use, change and share it, including at work. GitHub shows no license for the repository as a whole because each skill has its own.

Questions people ask

Is OpenAI's skills repository still maintained?

Its README has marked it deprecated since 2026-06-22. OpenAI now points to the openai/plugins repository for Codex skills and plugins. Skills in openai/skills still install as of 2026-10-01, but new work happens elsewhere.

Can I use OpenAI's PDF skill outside Codex?

Yes. Its folder has an Apache-2.0 license, and the skills CLI listed it for install on 2026-10-01. It only needs Python and Poppler, nothing specific to Codex.

  • DOCX: DOCX is Anthropic's proprietary agent skill for creating, reading and editing Word files, including tracked changes, comments and tables of contents; its license limits it to Claude and Anthropic's own products. (179,326 repository stars on October 2, 2026)
  • Obsidian Markdown: obsidian-markdown is Steph Ango's agent skill for writing Obsidian Flavored Markdown: notes with properties, wikilinks, embeds, callouts and block references that render correctly in an Obsidian vault. (49,077 repository stars on October 2, 2026)
  • PDF: PDF is Anthropic's proprietary agent skill for reading, creating and editing PDF files, from pulling out tables to filling forms and OCR; its license limits it to Claude and Anthropic's own products. (179,326 repository stars on October 2, 2026)
  • PPTX: PPTX is Anthropic's proprietary agent skill for building PowerPoint decks from scratch or from your template and checking every slide for overflow; its license limits it to Claude and Anthropic's products. (179,326 repository stars on October 2, 2026)

More from OpenAI

  • GitHub address comments: gh-address-comments is OpenAI's agent skill for pull request review feedback: the agent fetches every comment and review thread on the open PR, numbers them, and fixes only the ones you pick. (27,839 repository stars on October 2, 2026)
  • GitHub fix CI: gh-fix-ci is OpenAI's agent skill for failing pull request checks: the agent uses the GitHub CLI to find failing GitHub Actions runs, pulls the logs, summarizes the failure and proposes a fix plan. (27,839 repository stars on October 2, 2026)

Lists that include pdf