Inside ChatGPT’s New ‘Skills’: PDF & Spreadsheet Tools and Adding Them to Cursor

The author demonstrates that OpenAI has quietly integrated Anthropic‑style “Skills” into ChatGPT, exposing a /home/oai/skills directory with PDF and spreadsheet modules, explains how the PDF skill converts files to PNGs for vision‑based reading, and shows how to mount these skills in Cursor for local tool invocation.

AI Insight Log
AI Insight Log
AI Insight Log
Inside ChatGPT’s New ‘Skills’: PDF & Spreadsheet Tools and Adding Them to Cursor

Developer Simon Willison reported that OpenAI has quietly adopted Anthropic’s “Skills” mechanism, and the author set out to verify this claim using the GPT‑5.2 Thinking model, which can reveal its internal reasoning chain.

Testing the hypothesis

The author prompted the model with a request to create a PDF booklet on AI programming. During the model’s thinking process, a screenshot captured a command execution:

Executed bash -lc ls -R /home/oai/skills/pdfs.

Read a file named skill.md whose header was # PDF reading, creation, and review.

This indicated that the ChatGPT backend contains a pre‑populated /home/oai/skills directory holding standardized skill definitions.

Inspecting the PDF skill

By opening pdfs/skill.md, the author found the following guidance:

# PDF reading, creation, and review guidance

## Reading PDFs
- Use <code>pdftoppm -png $OUTDIR/$BASENAME.pdf $OUTDIR/$BASENAME</code> to convert PDFs to PNGs.
- Then open the PNGs and read the images.
- <code>pdfplumber</code> is also installed and can be used to read PDFs...
- Only do python printing as a last resort because you will miss important details...

The key takeaway is that ChatGPT now prefers to convert a PDF to PNG images and let a vision model “look” at the pages rather than extracting raw text.

PDF processing flow
PDF processing flow

Spreadsheet skill

The spreadsheets/skill.md file specifies that the AI should use Python libraries such as pandas or openpyxl to manipulate Excel data, ensuring accurate results instead of relying on the language model to guess.

Porting the skills to Cursor

Using a small prompt‑engineering trick, the author asked ChatGPT to package the entire /home/oai directory and send it. After downloading and extracting the archive, the directory structure showed folders like pdfs and spreadsheets, each containing a skill.md file.

Downloaded evidence
Downloaded evidence

To reuse these skills in the Cursor IDE, the author followed the same three‑step process described in a previous article:

Define the tool : write a script or Markdown describing the capability (e.g., “convert PDF”).

Mount the environment : place the files under .cursor/rules in the workspace.

Invoke directly : during a conversation, Cursor can call the local tool just like ChatGPT does.

Local file structure
Local file structure

In the author’s current AGENTS.md collection, skills such as pdf-to-md and youtube-transcript are already mounted, demonstrating that OpenAI’s hidden Skills mechanism is functional and can be leveraged in other developer tools.

Overall, the discovery confirms that OpenAI’s “Skills” are analogous to Anthropic’s approach, providing modular, command‑driven capabilities that enhance ChatGPT’s ability to handle PDFs, spreadsheets, and other file‑based tasks.

Original Source

Signed-in readers can open the original source through BestHub's protected redirect.

Sign in to view source
Republication Notice

This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactadmin@besthub.devand we will review it promptly.

Prompt engineeringChatGPTCursor IDESkillsAnthropicSpreadsheet automationPDF processing
AI Insight Log
Written by

AI Insight Log

Focused on sharing: AI programming | Agents | Tools

0 followers
Reader feedback

How this landed with the community

Sign in to like

Rate this article

Was this worth your time?

Sign in to rate
Discussion

0 Comments

Thoughtful readers leave field notes, pushback, and hard-won operational detail here.