Spreadsheets

Spreadsheets

Use the deterministic workbook CLI and keep calculations editable inside the resulting XLSX. Never overwrite an input workbook.

Runtime

Prefer workspace dependency Python. Otherwise use Python 3:

On macOS/Linux use python3; on Windows PowerShell use py -3 (or the workspace dependency loader's absolute python.exe path). On Windows set the skill directory with $env:SKILL_DIR = '<skill-directory>'. Every python3 example below uses this platform-specific launcher substitution.

python3 "$SKILL_DIR/scripts/bootstrap.py" <command> [arguments]

py -3 "$env:SKILL_DIR\scripts\bootstrap.py" <command> [arguments]

Missing packages are installed from the hashed lock into ~/.wegent-executor/plugin-envs/wework-public/spreadsheets/, never globally.

Workflow

1. Run inspect before modifying an existing workbook.
2. For new workbooks, prepare a JSON specification from spec.md, then run create.
3. Use formulas for auditable calculations. Use set-cell for focused changes and export-csv for interchange.
4. Run validate after every change.
5. Run render --images, inspect all pages, and correct clipped columns, excessive pagination, or unreadable formatting.

Commands

python3 "$SKILL_DIR/scripts/bootstrap.py" inspect --input source.xlsx --output inspection.json
python3 "$SKILL_DIR/scripts/bootstrap.py" create --spec workbook.json --output result.xlsx
python3 "$SKILL_DIR/scripts/bootstrap.py" set-cell --input source.xlsx --sheet Summary --cell B2 --value-json '125' --output result.xlsx
python3 "$SKILL_DIR/scripts/bootstrap.py" set-cell --input source.xlsx --sheet Summary --cell C2 --formula 'SUM(A2:B2)' --output result.xlsx
python3 "$SKILL_DIR/scripts/bootstrap.py" export-csv --input result.xlsx --sheet Summary --output summary.csv
python3 "$SKILL_DIR/scripts/bootstrap.py" validate --input result.xlsx
python3 "$SKILL_DIR/scripts/bootstrap.py" render --input result.xlsx --output-dir rendered --images

Quality rules

• Store numbers, dates, booleans, and formulas as typed cell values rather than formatted strings.
• Use a styled header, readable column widths, freeze panes, and filters for non-trivial tables.
• Prefer formulas over hard-coded calculated outputs when users need an auditable model.
• Remember that the Python library does not calculate formulas; recalculate with Excel or LibreOffice when current values matter.
• Validate structure and visually review rendered output before delivery.