agent-xlsx
A command-line tool for AI Agents to deeply analyze Excel files
Install & Use
Copy this prompt and send it to your AI assistant (Claude / Cursor / TRAE / Codex / WorkBuddy etc.) to auto-install:
Help me install this AI Skill: agent-xlsx. It is used for: A command-line tool for AI Agents to deeply analyze Excel files Full Skill content: https://321skill.com/skills/agent-xlsx/raw/index.md Read that page and install it.
The prompt includes a link to the full Skill content. You can also view the full content.
This skill addresses the pain points of low efficiency and token wastage when AI Agents process Excel files. In practical development, an agent needs to understand a workbook's structure, data, formulas, charts, and VBA like a human would. However, traditional libraries (e.g., openpyxl) are designed for humans writing scripts—reading a sheet once consumes a large number of tokens and does not allow for progressive information retrieval. Following three principles—progressive disclosure, speed first, and token efficiency—agent-xlsx enables an AI Agent to fully grasp the content of an Excel file within 1-2 rounds of interaction.
Usage is straightforward: simply install agent-xlsx in your terminal and execute commands via the command line, such as probe (to explore structure), read (to read data), search (to search content), screenshot (to capture screenshots), and inspect (to examine metadata). You can then instruct the AI Agent using natural language to call these commands. For example: first run probe to see what worksheets the workbook contains, then use read to retrieve specific areas as needed. All output is in JSON format, which is extremely token-efficient.
It is well-suited for data analysis teams, agent developers, and enterprise automation scenarios that frequently handle Excel reports. Particularly for users already employing AI coding assistants like Claude or Cursor who want their AI to understand Excel workbooks layer by layer like a human, this tool can quickly set up data processing pipelines and reduce manual intervention.
We recommend using this tool uniformly when batch processing Excel files or building automated reporting workflows. Note that it primarily relies on Aspose.Cells (commercial license) for screenshot and recalculation features; without a license, watermarks and a limit of 100 files per session apply. If screenshots are not needed, all other commands have zero system dependencies and can be used directly.
Key Features
Unlike traditional Excel libraries such as openpyxl and xlrd, agent-xlsx is built on Polars + fastexcel (Rust/Calamine), offering 7-10x faster read speeds. Its output is optimized for token efficiency and supports progressive disclosure (probe → read → inspect), whereas traditional libraries require loading all data at once, resulting in high token consumption and no on-demand retrieval capability.
Limitations
Requires Python 3.8+ environment. Screenshot and recalculation features depend on Aspose.Cells (commercial license; watermarks and a limit of 100 files per session apply without a license) or LibreOffice (optional system dependency). All other commands (probe, read, search, etc.) have no additional dependencies.
FAQ
How do I install agent-xlsx?
We recommend using `pip install agent-xlsx`. You can also use `uv tool install agent-xlsx` or `pipx install agent-xlsx`. For temporary runs, use `uvx agent-xlsx`.
Does it support reading formulas and macros in Excel?
Yes. The `probe` and `read` commands can extract formula text, the `vba` subcommand can extract VBA macro code, and `screenshot` can render cell results containing formulas.
Installation guide for AI assistants
If your AI coding assistant (Claude Code, Cursor, TRAE etc.) can see this page, send it this message to auto-install:
Visit https://321skill.com/skills/agent-xlsx/raw/index.md to read the original Skill definition (Markdown format) for agent-xlsx, and install it according to the instructions.
Raw Markdown URL for AI: /skills/agent-xlsx/raw/index.md