GitHub radar
huashu-excel: Error-Proof Excel Analysis Skills for AI Agents
A skill for an AI agent to analyze excel files, clean and validate the data, then calculate the sum of the values and compare it with the subtotals in the excel file, showing any difference or error.
huashu-excel is a Claude Code/Cursor/Codex skill which can be installed on any agents that support skills format. This skill will equip your agent with an entire workflow of excel analysis, including a preliminary structured diagnosis of the excel file to identify issues such as header error, merged cells, subtotal rows in the body, duplicate entries, numeric values presented as text, and different date formatting. Then the skill will clean the data and compare the calculated sum with the subtotals in the excel file. Any difference will be reported to you with clear explanation and the final result will be withheld until you understand why. For example, when analyzing this Chinese sales data, using the traditional pd.read_excel() method leads to a total that is 161% higher than the actual value without any warning or error. But with huashu-excel, all the three issues (subtotal row mistakenly included as one store, double counting of regional subtotal, duplicate entry) are identified in the diagnosis phase. This skill requires only openpyxl as external package.
Why a vibe-coder should care
Now, your agent will not only give you a result, but also tell you how it got there, so you can share the result with others confidently without having to worry about the “how?” question. Particularly suitable for sales data, financial statements, and other excel files that have been handled by multiple people.
How to install
Copy this and send it to your agent — Claude Code, Codex, any of them:
Install the skill from https://github.com/alchaincyf/huashu-excel using npx skills add, then take my Excel file and run the full pipeline: health-check, cleaning, total reconciliation, and a final report.
Runs on a regular laptop — only Python needed.
Open on GitHub▌ More finds