The Claude Skills Library.
A curated directory of Claude Skills — capabilities that extend Claude for specific workflows, roles, and outputs. From Anthropic-official skills to community-submitted, all reviewed and vetted by AI ARMY before listing.
Official Anthropic Skills
Generate, edit, fill, and extract from PDF files programmatically. The most-used skill in the library.
DOCX
Create, read, and edit Microsoft Word documents. Preserves styles, headings, tables, tracked changes.
XLSX
Create, edit, and parse Excel spreadsheets with formulas, formatting, and pivot tables intact.
PPTX
Generate and edit PowerPoint presentations. Slide layouts, charts, speaker notes, branded templates.
Brand Guidelines
Apply your brand colors, fonts, voice rules, and logo placement consistently across all Claude outputs.
Canvas Design
Compose visuals on a 2D canvas with layers, typography, and spacing. The Canva/Figma primitives.
Theme Factory
Generate complete color palettes, typography sets, and component styles for product or content design.
Algorithmic Art
Create parametric and procedural artwork via p5.js with seeded randomness and interactive parameters.
Claude API
Teaches Claude the up-to-date patterns for calling the Anthropic API — auth, streaming, tools, vision, caching.
MCP Builder
Scaffold MCP servers and tools. Critical for anyone building agentic workflows in 2026.
Frontend Design
Build production-grade frontend components and pages with strong visual design principles. React, HTML, CSS, Tailwind.
Web Artifacts Builder
Generate self-contained single-file HTML/JS artifacts. Run anywhere without a build step.
Webapp Testing
Write and run Playwright or Puppeteer tests headlessly. Closes the 'generated code but didn't verify' gap.
Doc Coauthoring
Three-phase workflow for collaborating on long-form documents: gather → refine → reader test.
Internal Comms
Standardize format and tone of all-hands updates, status reports, announcements, internal memos.
Slack GIF Creator
Generate animated reaction GIFs sized for Slack reactions and inline replies. Themed presets included.
Skill Creator
The meta-skill: walks Claude through generating new custom skills in the standard SKILL.md format.
Anthropic Vertical Collections
Anthropic's official vertical packs — each ships multiple skills, agents, and MCP connectors targeted at a specific function. Knowledge Work Plugins is a strong starting point for each business role; AI ARMY Skill Packs (above) go further with multi-source orchestration vetted for production.
Claude for Financial Services
Pitch Agent — comps, precedents, and LBO → a branded pitch deck, end to end. + 10 more agents (Market Researcher, Earnings Reviewer, Model Builder, Valuation Reviewer, GL Reconciler, Month-End Closer, Statement Auditor, KYC Screener, Meeting Prep) and skill bundles for investment-banking, equity-research, private-equity, wealth-management, and fund-admin.
Claude for Healthcare
prior-auth — reviews prior-authorization requests with clinical-documentation synthesis (payer/provider). + clinical-trial-protocol (pharma, FDA/NIH-compliant) and fhir-developer (FHIR R4 / SMART auth). Ships with five hosted MCP servers: CMS Coverage, ICD-10, NPI Registry, Clinical Trials, PubMed.
Claude for Legal
Vendor Agreement Reviewer — reviews a vendor MSA against your playbook, produces a redline memo. + NDA Triager (GREEN/YELLOW/RED triage), Amendment Tracer, and scheduled watchers (Renewal Watcher, Deal Debrief, Playbook Monitor). MCP connectors include Ironclad, DocuSign, iManage, CourtListener. Drafts only — every output is for attorney review, not legal advice.
Defending Code Reference Harness
/threat-model — build a threat model, then run a static scan + triage. + /quickstart, /vuln-scan, /triage, /patch, /customize, and an autonomous reference pipeline (recon → find → verify → triage → patch). Skills only read/write within your own repo (safe-by-scope). Anthropic frames this as a reference harness you customize, not a turnkey pack.
Launch Your Agent
launch-your-agent — interviews you, scopes a v0, launches in your account on Claude Managed Agents, grades against your definition of done, iterates, and schedules it. + wrap-up companion skill for close-out and status. 4-phase flow: interview → stage & launch → grade & iterate → run without you. Highly relevant to a founder/builder audience.
Knowledge Work Plugins
Role-based official plugins — Sales, Marketing, Finance, Data, Customer-Support, Product-Management, Legal, Enterprise-Search, Bio-Research, Productivity, and Cowork-Plugin-Management. Each bundles skills + connectors + slash commands + sub-agents across Slack, HubSpot, Notion, Linear, Jira, Snowflake, Databricks, Canva, Figma, and more. Anthropic frames these as a "strong starting point" you customize to your tools and terminology.
Partner Skills
Official partner-built skills — vetted by Anthropic — that teach Claude to work within the tools you already use, not just connect to them. The full directory grows; the row below is our editorial pick.
Notion
Lets Claude read pages, databases, and comments inside your Notion workspace — and write back updates with full citation.
Canva
Generates and edits Canva designs from a Claude conversation — slides, social posts, brand-templated layouts.
Figma
Reads frames, components, and tokens; drafts specs and code-from-design handoffs without leaving Claude.
Atlassian (Jira / Confluence / Trello)
Reads tickets, docs, and boards; drafts issue updates, sprint summaries, and runbooks across the Atlassian suite.
Cloudflare
Inspects Workers, KV, R2, D1, and Hyperdrive configs; drafts edge logic and surfaces production issues from logs.
Sentry
Reads errors and performance events, summarizes regressions, and drafts fix PRs grounded in the failing stack.
Ramp
Reads spend, cards, and bills; drafts variance notes, vendor consolidations, and accrual entries from real ledger data.
Stripe
Reads charges, customers, subscriptions, and disputes; drafts refunds, dunning logic, and revenue summaries with audit trail.
Zapier
Triggers Zaps and reads workflow run history — connects Claude to 7,000+ apps via the existing Zapier integration mesh.
Community Skills
Community Skills — Vetted, Coming Soon
The open skills ecosystem is enormous — millions of public skills, largely unaudited. Independent audits have found prompt-injection and quality issues in a large share of community skills. We're personally reviewing external skills for safety and security before listing any here. When a community skill earns a place in this library, it's because we checked it.
Built a Claude Skill?
Submit it for review. We vet every submission before adding to the library — typically within 7 days.