Find your symptom below. If a file of yours breaks a script, Claude shows the error; the fixes below cover the common causes.
| Symptom | Fix |
|---|---|
| A store export is not recognised | Use the column map described in the store-export-normalizer skill to name your columns once. |
| Margins look too high or too low | Set your own fee rules and costs in the editable fee file; the defaults are placeholders. |
| Claude answers without running a script | Check the skill is toggled on under Customize > Skills and code execution is on. Then ask again and name the job, e.g. "use the sales-anomaly-watch skill". |
ModuleNotFoundError: pandas in Claude Code |
Run pip install pandas openpyxl in the Python that Claude Code uses. Claude.ai and the API already have them. |
| No PNG charts in the output | matplotlib is missing. Everything else still works; install matplotlib if you want charts. |
| Upload fails in claude.ai | Upload the small zips inside claude-ai-uploads/, one per skill, not the whole pack zip. |
python tests/run_tests.py shows fewer than 33/33 |
Run it from the pack folder with Python 3.9+, pandas and openpyxl installed. If it still fails, compare tests/run_log.txt with TESTS.md. |
FAQ
Does it connect to my store? No. It reads the exports you download, so there are no logins or API keys to hand over.
Is margin the same as profit? No. It is contribution margin, before ads, salaries and overheads. The reports say so.
Do I need to code? No. You upload the skills once, then attach files in a chat and ask in plain words.
Does it send my data anywhere? No. The scripts make no network calls, and customer emails are hashed in the clean file.
Which Claude plans work? Any plan where Skills and code execution are on (Pro, Max, Team, Enterprise), Claude Code, or the Claude API with the code-execution tool.
Where do I get help? This file and README.md. There is no call, chat or email support. Fixes ship as new versions at your download link; CHANGELOG.md lists them.