| name | documents |
| description | Read and produce documents — parse xlsx, docx, pptx, and pdf to extract data, and generate polished workbooks, memos, reports, and PDFs. Use when someone uploads or references a spreadsheet, Word document, slide deck, or PDF (invoices, billing workbooks, contracts, court filings, financial reports), or asks to read, extract, summarize, convert, or produce one. |
Documents
The firm lives in spreadsheets, PDFs, and Word docs — invoices, the monthly billing workbook, contracts, court filings, financial reports. This skill reads them for data and produces clean ones back.
Reading / extracting
- xlsx — the billing workbook, rate sheets, financial reports, contact/matter exports. Pull the rows you need and keep structure.
- pdf — invoices (for billing reconciliation), court orders, recorded documents, financial statements. Extract figures and line items; OCR scanned pages when needed.
- docx — engagement letters, motions, memos, receiver reports.
- pptx — decks and exhibits.
When the host environment provides dedicated xlsx / docx / pptx / pdf skills, use them for the mechanics — this skill is about why and what: which numbers to pull, how they feed billing, what to summarize.
Formats & token-optimized parsing
Read and create pdf, docx, xlsx, csv (and pptx).
- Token-optimized reading: convert documents to compact Markdown/text first — markitdown for pdf/docx/pptx,
openpyxl/csv for spreadsheets — and pull only the rows/sections you need. Never load a whole workbook or a 25k-row export into context: filter or aggregate a large CSV (e.g. the activities export) with a small script and read the result, not the raw file.
- OCR scanned invoices/orders only when plain text extraction comes back empty.
- Creating: emit the Clio import csv (per
billing-format.md — every field quoted, UTF-8 BOM), the billing xlsx workbook, and pdf/docx memos (per output-style.md).
Producing
Generate the billing CSV, the polished billing workbook (.xlsx), a month-end memo (.pdf/.docx), a reconciliation report, or a summary — to the layout and styling in ${CLAUDE_PLUGIN_ROOT}/references/output-style.md and the schema in ${CLAUDE_PLUGIN_ROOT}/references/billing-format.md. Use the host xlsx, pdf, and docx skills for the mechanics; this skill decides what goes in. Keep every deliverable firm-grade: restrained, legible, court-facing — never a raw dump.
Care
- Treat document contents as sensitive — many are privileged or contain PII (see
billing-rules). Describe and extract what's needed; don't surface privileged substance unnecessarily.
- Never modify the firm's original/source files — produce new outputs.
- This works the same in Claude.ai, Cowork, and Claude Code.
