
tfriedel/claude-office-skills
4 skills5.6k installs2.9k starsGitHub
Install
npx skills add https://github.com/tfriedel/claude-office-skillsSkills in this repo
1Xlsxxlsx is an agent skill from the claude-office-skills collection for working with Microsoft Excel-compatible spreadsheets. It targets solo and indie builders who need reliable .xlsx handling inside Claude Code, Cursor, Codex, or similar agents—especially when formulas must evaluate correctly after programmatic edits. The bundled workflow loads workbooks with openpyxl, ensures a LibreOffice RecalculateAndSave macro exists, and runs LibreOffice to calculate and persist results. Use it when you are generating financial models, importers, roster exports, or fixing agent-written sheets that show formula text instead of computed values. It fits the Build phase when spreadsheets are deliverables, and also Validate and Operate when pricing tables or ops reports live in Excel. It is a procedural office integration, not a hosted API; you need LibreOffice and Python dependencies locally.2.2kinstalls2Docxdocx is an agent skill that teaches coding agents to build Microsoft Word .docx files with the JavaScript docx library instead of guessing XML or markdown exports. Solo builders use it when contracts, reports, one-pagers, or customer deliverables must be real DOCX with tables, TOC, headers, and footnotes. The tutorial is procedural: read the whole skill before generating, install docx, compose Document sections, and save via Packer. It stresses common failure modes—line breaks via separate Paragraphs, shading and border enums, hyperlink types—so agents stop shipping files Word cannot open cleanly. The same patterns work in Node automation and browser download flows. It complements markdown docs skills by targeting stakeholders who require editable Word, and it spans early specs in Validate through shipped collateral in Launch when you still author in the repo.1.4kinstalls3Pptxpptx is an agent skill for solo builders who want PowerPoint output without rebuilding slides by hand in Office. It centers on converting HTML slides through the html2pptx.js library so typography, lists, images, and placeholder regions land in the correct coordinates on standard aspect ratios. The guide is opinionated about structure: every line of text must sit inside semantic text tags, manual bullet characters and br-based line breaks are forbidden, and layout dimensions are fixed in points so agents do not guess page size. That makes it especially useful when an agent already drafted slide HTML during planning or demo prep and you need a file investors, customers, or teammates can open offline. It is narrower than a full design system—it does not replace narrative strategy—but it closes the gap between web-native drafts and shareable decks.1.4kinstalls4Pdfpdf is an Office & Documents agent skill from tfriedel/claude-office-skills that teaches your agent how to handle real-world PDF jobs without guessing libraries or coordinates. Solo and indie builders install it when contracts, tax forms, spec PDFs, or lead-magnet downloads need reliable filling, extraction, or assembly inside Claude Code, Cursor, or Codex—not in a separate desktop app. The workflow is deliberately procedural: for forms you must detect fillable AcroForm fields before choosing the pypdf JSON path or the rasterized annotation path with pdf2image and bounding boxes. Beyond forms, the skill documents merge and split with pypdf, layout-aware text and table pulls with pdfplumber, optional Camelot and tabula fallbacks, OCR for scanned pages, and creation paths through ReportLab Platypus, Markdown or HTML via pandoc, or headless LibreOffice. It matters because PDF coordinate systems and field types trip up one-shot coding; this packages battle-tested scripts and guardrails so agents deliver verifiable files in validate, build, launch, and operate work.585installs