Create Spreadsheet

Documents

create-spreadsheet

Create Spreadsheet is an Agent Skill for documents published in the theexperiencecompany/gaia repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Generate an Excel (.xlsx) workbook or a CSV file — tables, financial models, data exports, formatted reports with charts. Use when the user wants a spreadsheet or CSV. Built with openpyxl + pandas (Python).

View on GitHub ↗

QUICK START

How to use this skill

Bring this guide into your coding agent with a prompt tailored to the tool you use.

Open your project in Codex.
Copy and paste the prompt below.
Review files and risks before approving any action.

Prompt to paste

Inspect the Agent Skill at https://github.com/theexperiencecompany/gaia/blob/HEAD/apps/api/app/agents/skills/builtin/create-spreadsheet/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/create-spreadsheet, but wait for my approval before writing files or running scripts.

Copying this prompt does not install or run the skill. Review third-party files before use.

Create Spreadsheet

When to use

The deliverable is .xlsx or .csv. For Google Sheets use the googlesheets skill.

Important: values, not live formulas

There is no spreadsheet recalc engine in the sandbox. If you write a formula like =SUM(...), Excel shows it as blank until the user opens and recalculates. So:

Compute results in Python (pandas/numpy) and write the final values. This is the default and it always opens correctly.
Only add a literal =FORMULA when the user explicitly wants an editable formula — and when you do, also write the computed value in an adjacent labeled cell so the number is visible immediately.

How it works

A template here is a Python program (openpyxl + pandas) that writes the workbook. You adapt the data and structure, then the build script runs it and validates the file.

Skill directory: /workspace/integrations/docgen/agent/skills/create-spreadsheet — templates/, reference.md, and scripts/ live here; read them by absolute path. Your work goes in ./scratch/.

Workflow (.xlsx)

Work in a job dir: mkdir -p ./scratch/xlsx-<short-name>.
Adapt the template — don't start blank. read /workspace/integrations/docgen/agent/skills/create-spreadsheet/templates/report.py, then write the edited program to ./scratch/xlsx-<name>/sheet.py. Read /workspace/integrations/docgen/agent/skills/create-spreadsheet/reference.md for openpyxl building blocks and the values rule.

Build + validate:

bash /workspace/integrations/docgen/agent/skills/create-spreadsheet/scripts/build.sh ./scratch/xlsx-<name>/sheet.py ./scratch/xlsx-<name>/out.xlsx

Prints OK: <path> (sheets=N) or ERROR: ....

Fix loop: on ERROR, read the message, fix, re-run. Cap at 5 attempts.
Deliver: mv ./scratch/xlsx-<name>/out.xlsx ./artifacts/<final-name>.xlsx, then report the full path.

CSV (no build needed)

CSV is plain text — write it directly with bash and deliver:

python3 - <<'PY'
import csv
rows = [["name","amount"], ["Acme", 1200], ["Globex", 980]]
with open("./artifacts/<final-name>.csv", "w", newline="") as f:
    csv.writer(f).writerows(rows)
PY

No toolchain, no validation step — just confirm the file is in ./artifacts/ and report the path.

Toolchain note

The build script provisions openpyxl + pandas on first use; the first workbook may take a little longer, then it's fast.

Never

Never write = formulas as the only representation of a computed number (no recalc → blank cells). Write the value.
Never leave the file in ./scratch/. Never skip build/validate for .xlsx.

Templates

report.py — formatted workbook: styled header, data rows, a computed totals row, number formats, column widths, and a bar chart.