Office Xlsx

Documents

office-xlsx

Office Xlsx is an Agent Skill for documents published in the shiwenwen/hope-agent repository on GitHub: a SKILL.md instruction file an AI agent can load to do this task.

Use when the user asks to create, inspect, verify, analyze, format, or deliver Excel `.xlsx` workbooks, Google Sheets-targeted spreadsheet artifacts, trackers, budgets, models, tables, dashboards, formulas, CSV/TSV-to-XLSX conversions, or spreadsheet-ready data packs.

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/shiwenwen/hope-agent/blob/HEAD/skills/office-xlsx/SKILL.md and all supporting files. Report provenance, permissions, scripts, dependencies, and security risks. Propose an exact installation into .agents/skills/office-xlsx, 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.

Office XLSX

Use the bundled scripts in this skill package to produce editable .xlsx workbooks. The builder writes formulas, styles, bar/column/line/pie charts, real Excel tables, data validation, conditional formatting, and workbook recalculation hints. The skill activation metadata includes Skill directory; treat that as SKILL_DIR and run scripts from SKILL_DIR/scripts/.

Workbook Shape

For nontrivial workbooks, prefer:

Summary or dashboard sheet first.
Inputs / assumptions sheet next.
Detail or source data sheets after that.
Checks sheet only when formulas, reconciliations, or model integrity matter.

Workflow

Normalize source data before writing the workbook.
Use formulas for derived values instead of hardcoded calculated outputs. Strings beginning with = are written as Excel formulas.
For structured workbook creation, create a JSON spec in the working directory and run:

python3 "$SKILL_DIR/scripts/check_env.py"
python3 "$SKILL_DIR/scripts/build_xlsx.py" --spec spec.json --out output.xlsx
python3 "$SKILL_DIR/scripts/inspect_xlsx.py" --verify output.xlsx
python3 "$SKILL_DIR/scripts/formula_audit.py" output.xlsx

For CSV/TSV conversion, run one or more inputs into one workbook:

python3 "$SKILL_DIR/scripts/csv_to_xlsx.py" --input data.csv --input lookup.tsv --sheet Data --sheet Lookup --out output.xlsx
python3 "$SKILL_DIR/scripts/inspect_xlsx.py" --verify output.xlsx

If visual QA matters and LibreOffice is available, run:

python3 "$SKILL_DIR/scripts/render_preview.py" output.xlsx

To patch an existing workbook without rebuilding it, create a patch JSON and run:

python3 "$SKILL_DIR/scripts/patch_xlsx.py" --input existing.xlsx --patch patch.json --out output.xlsx
python3 "$SKILL_DIR/scripts/inspect_xlsx.py" --verify output.xlsx
python3 "$SKILL_DIR/scripts/formula_audit.py" output.xlsx --write-cache cached-output.xlsx

Patch actions:

{
  "actions": [
    {"action": "append_rows", "sheet": "Data", "rows": [["New", 123, "=B2*2"]]},
    {"action": "set_cell", "sheet": "Summary", "cell": "B2", "value": "=SUM(Data!B:B)"}
  ]
}

When formulas matter, use formula_audit.py --write-cache and deliver the cached workbook unless the audit reports unsupported formulas that require Excel/LibreOffice recalculation. For broad formula coverage, run:

python3 "$SKILL_DIR/scripts/recalculate_xlsx.py" output.xlsx --out recalculated.xlsx
python3 "$SKILL_DIR/scripts/inspect_xlsx.py" --verify recalculated.xlsx

Deliver the .xlsx path or attach it with send_attachment.

Spec Shape

{
  "title": "Workbook title",
  "sheets": [
    {
      "name": "Summary",
      "rows": [["Metric", "Value"], ["Revenue", 1200000], ["Margin", "=B2*0.42"]],
      "tables": [{"name": "SummaryTable", "ref": "A1:B3"}],
      "data_validations": [{"range": "A2:A10", "type": "list", "formula1": ["Revenue", "Margin"]}],
      "conditional_formats": [{"range": "B2:B10", "type": "colorScale"}],
      "charts": [
        {"type": "column", "title": "Summary", "categories": "$A$2:$A$3", "values": "$B$2:$B$3", "anchor": "D2"},
        {"type": "line", "title": "Trend", "categories": "$A$2:$A$3", "values": "$B$2:$B$3", "anchor": "D18"}
      ],
      "column_formats": ["text", "currency"],
      "column_widths": [24, 16],
      "freeze_top_row": true,
      "autofilter": true
    }
  ]
}

Quality Bar

Keep important values visible; avoid tiny columns and clipped headers.
Use one workbook, not many disconnected CSV-like sheets, when relationships between tabs matter.
Prefer real Excel tables, filters, validations, conditional formats, and charts when the workbook is meant to be used repeatedly.
Keep formulas editable, audit formulas before delivery, and write cached values for supported formulas when possible. Supported audit functions include common arithmetic, comparisons, SUM, AVERAGE, MIN, MAX, COUNT, MEDIAN, ROUND, ABS, and simple IF.
When patching an existing workbook, preserve unrelated package parts and rerun inspect_xlsx.py plus formula_audit.py; use recalculate_xlsx.py when the formula surface exceeds the bundled evaluator.
If preview rendering fails because LibreOffice or a PDF-to-PNG renderer is missing, state exactly which verification passed; do not imply visual QA passed.