Reports from tables
Use it with your AI
Works in Claude, ChatGPT or any other AI
-
Step 2: Ask your AI
Paste this into your chat:
What does the Pacpio project pacpio-team/reports-from-tables do?
- data (folder) 1 file · 224 B
- knowledge (folder) 3 files · 4.4 KB
- licenses (folder) 1 file · 1.1 KB
- my (folder) 2 files · 507 B
- reports (folder) 0 files · 0 B
- skills (folder) 5 files · 7.8 KB
- templates (folder) 1 file · 1.1 KB
- AGENTS.md 2.6 KB
- README.md 2.4 KB
15 files · 20 KB
Reports from tables
For people who have a table but no analyst: a sales export, expenses, attendance, orders, sign-ups. This project describes the table, finds problems in the data, builds a report by one template every period, compares periods, and keeps the tables and the reports together. Every number comes with how it was counted, so you can check it.
How to start
- Make a copy of this project (the "Make a copy" button).
- Connect your AI to Pacpio if you have not yet.
- Ask: "Here is my sales table for September, make a monthly report", and paste the table.
The first time, your assistant asks three questions: currency and units, which numbers and breakdowns matter, who reads the reports. It saves the answers in my/report-settings.md.
What is inside
AGENTS.md— how your assistant works here: describe, check, then report; parts must add up; no causes, only what the data shows; source files are never changed.skills/— the five things it does:save-table— a pasted table into a clean CSV file indata/.describe-table— columns, types, period, missing values, totals.data-check— problems to fix before counting, with row numbers.report— the report by the template, with "How this was counted".compare-periods— what grew and what fell between two periods.
templates/report.md— the shape of every report.knowledge/— the counting rules, the report rules, a worked example.my/report-settings.md— your units, metrics, breakdowns and reader.data/— your tables, one CSV per table, with an index indata/README.md.reports/— profiles, checks, reports and comparisons.
Things you can ask
- "Save this table as September sales."
- "What is in this file, and is anything wrong with it?"
- "Monthly report, breakdown by region."
- "Compare September with August."
- "Which rows look odd?"
Tips
- Spreadsheet files are not read here yet: paste the sheet as text, or save it as CSV first.
- For a big table, paste the totals from your spreadsheet as well. The assistant uses them as a check.
- Read "How this was counted" at the end of each report. If a number matters, redo it once yourself.
Credits and licenses
The CSV format rules are adapted from the pattern export_data_as_csv from Fabric by Daniel Miessler and contributors, MIT License (licenses/fabric-MIT.txt). The rest is written for this project.