Skills & Plugins
Skills are modular instruction packages that extend Claude Code with focused workflows, domain expertise, and reusable automation. They load only when needed — keeping your context lean while giving Claude deep capabilities on demand.
Why Skills Instead of CLAUDE.md?
| Aspect | CLAUDE.md | Skills |
|---|---|---|
| When loaded | Every message (constant overhead) | On-demand only |
| Token cost | Always consuming context | Only when active |
| Best for | Project context, conventions | Focused workflows |
| Can execute code | No (instructions only) | Yes (via scripts/) |
How Skills Work
Skills live in your project or personal directory:
your-project/.claude/skills/<skill-name>/SKILL.md # Project-level
~/.claude/skills/<skill-name>/SKILL.md # Personal (all projects)
Every skill requires a SKILL.md file with YAML frontmatter:
---
name: my-skill
description: Short description of what this skill does
trigger: /my-skill
---
# Skill Instructions
Detailed instructions Claude follows when activated.
| Invocation | Example | When |
|---|---|---|
| Slash command | /pdf, /commit | User explicitly types |
| Auto-trigger | Claude detects match | Skill description matches intent |
Anthropic Official Skills (17 skills)
Source: github.com/anthropics/skills
Document & File Skills:
| Skill | Command | What It Does |
|---|---|---|
/pdf | Extract text/tables, create PDFs, merge/split, OCR. Includes 12 Python scripts. | |
| docx | /docx | Word document creation, editing, tracked changes |
| pptx | /pptx | Presentation creation, editing, templates |
| xlsx | /xlsx | Spreadsheet creation, formulas, formatting, charts |
Design & Frontend Skills:
| Skill | Command | What It Does |
|---|---|---|
| frontend-design | /frontend-design | Production-grade frontend UI with high design quality |
| algorithmic-art | /algorithmic-art | p5.js generative art with seeded randomness |
| canvas-design | /canvas-design | Visual art in PNG/PDF — posters, diagrams. 78+ fonts. |
| web-artifacts-builder | /web-artifacts-builder | Multi-component React + Tailwind HTML artifacts |
Development & Testing Skills:
| Skill | Command | What It Does |
|---|---|---|
| webapp-testing | /webapp-testing | Playwright-based web app testing and debugging |
| mcp-builder | /mcp-builder | Guide for creating MCP servers in Python and TypeScript |
| skill-creator | /skill-creator | Guide for creating new Claude Code skills |
Communication & Branding Skills:
| Skill | Command | What It Does |
|---|---|---|
| brand-guidelines | /brand-guidelines | Anthropic’s official brand colors and typography |
| theme-factory | /theme-factory | 10 professional themes for slides, docs, reports |
| internal-comms | /internal-comms | Internal communications — updates, newsletters, FAQs |
| doc-coauthoring | /doc-coauthoring | Structured documentation co-authoring workflow |
| slack-gif-creator | /slack-gif-creator | Animated GIFs optimized for Slack |
| keybindings-help | /keybindings-help | Customize keyboard shortcuts |
Knowledge Work Plugins (47 skills across 10 domains)
Source: github.com/anthropics/knowledge-work-plugins | Apache-2.0
Most have auto-trigger enabled — Claude activates them automatically when your request matches the domain.
Productivity (2 skills): Memory management (workplace shorthand, people, projects), task management (TASKS.md tracking). Commands: /productivity:start, /productivity:update
Sales (6 skills): Account research, call prep, daily briefing, draft outreach, competitive intelligence, asset creation. Commands: /sales:call-summary, /sales:forecast, /sales:pipeline-review
Finance (6 skills): Journal entries, reconciliation, financial statements, variance analysis, audit support, close management. Commands: /finance:journal-entry, /finance:reconciliation, /finance:income-statement, /finance:variance-analysis, /finance:sox-testing
Data (7 skills): SQL queries, data exploration, visualization, statistical analysis, data validation, interactive dashboards, context extraction. Commands: /data:write-query, /data:explore-data, /data:create-viz, /data:analyze, /data:validate, /data:build-dashboard
Legal (6 skills): Contract review, NDA triage, compliance (GDPR/CCPA), canned responses, risk assessment, meeting briefing. Commands: /legal:review-contract, /legal:triage-nda, /legal:vendor-check, /legal:brief, /legal:respond
Marketing (5 skills): Content creation, campaign planning, brand voice, competitive analysis, performance analytics. Commands: /marketing:draft-content, /marketing:campaign-plan, /marketing:brand-review, /marketing:competitive-brief, /marketing:email-sequence, /marketing:seo-audit
Customer Support (5 skills): Ticket triage, response drafting, escalation, knowledge management, customer research. Commands: /customer-support:triage, /customer-support:draft-response, /customer-support:escalate, /customer-support:kb-article
Product Management (6 skills): Feature specs/PRDs, roadmap management, competitive analysis, metrics tracking, stakeholder comms, user research synthesis. Commands: /product-management:write-spec, /product-management:roadmap-update, /product-management:metrics-review
Enterprise Search (3 skills): Search strategy, knowledge synthesis, source management. Commands: /enterprise-search:search, /enterprise-search:digest
Plugin Management (1 skill): Create and customize plugins for your organization.
BMAD Method (9 agents + 15 commands)
Source: github.com/bmad-code-org/BMAD-METHOD
BMAD (Breakthrough Method for Agile AI-Driven Development) is a full product lifecycle framework with 9 specialized AI agent personas:
| Agent | Role |
|---|---|
| BMad Master | Core orchestrator, workflow routing |
| Business Analyst | Product brief, requirements gathering |
| Product Manager | PRDs, tech specs |
| System Architect | Architecture design, solutioning gates |
| UX Designer | UX design creation |
| Scrum Master | User stories, sprint planning |
| Developer | Story implementation |
| Creative Intelligence | Brainstorming, research |
| Builder | Create custom agents and workflows |
Key commands: /bmad:workflow-init, /bmad:prd, /bmad:tech-spec, /bmad:architecture, /bmad:create-story, /bmad:sprint-planning, /bmad:dev-story
Community Skills
| Resource | Description |
|---|---|
| Awesome Claude Skills | Community-curated list (5.9k+ stars) |
| Claude Office Skills | Office document skills |
| Claude Skills Collection | Curated official + community skills |
Community skills are not reviewed by Anthropic. Always inspect a skill’s SKILL.md and scripts before installing.
Installing Skills
git clone https://github.com/anthropics/skills.git
cp -r skills/skills/pdf ~/.claude/skills/
cp -r skills/skills/docx ~/.claude/skills/
cp -r skills/skills/pptx ~/.claude/skills/
git clone https://github.com/anthropics/knowledge-work-plugins.git
cp -r knowledge-work-plugins/plugins/*/skills/* ~/.claude/skills/
git clone https://github.com/aj-geddes/claude-code-bmad-skills.git
cp -r claude-code-bmad-skills/skills/bmad ~/.claude/skills/
Creating Your Own Skills
Method 1: Ask Claude
Create a skill file for our deployment workflow. It should:
1. Check all tests pass
2. Build the Docker image
3. Push to our registry
4. Update the Kubernetes deployment
Save it to .claude/skills/deploy/SKILL.md
Method 2: Use the Skill Creator
/skill-creator to launch the built-in skill creation guide.Method 3: Manual Creation
Create ~/.claude/skills/my-skill/SKILL.md:
---
name: my-skill
description: What this skill does
trigger: /my-skill
---
# My Skill
## When Triggered
1. Step one
2. Step two
## Rules
- Rule 1
- Rule 2
Skills vs CLAUDE.md vs MCP — When to Use What
| Aspect | CLAUDE.md | Skills | MCP Servers |
|---|---|---|---|
| What it is | Project instructions | Modular instruction packages | External service connectors |
| When loaded | Every message | On-demand | Always active (when enabled) |
| Token cost | Constant overhead | Only when active | Tool definitions always in context |
| Best for | Project context, conventions | Focused workflows, domain expertise | External data, APIs, databases |
| Can execute code | No | Yes (via scripts/) | Yes (server-side) |
| Portable | Text file, any editor | Claude Code only | Open protocol, multi-client |
Rule of thumb: Put what Claude always needs to know in CLAUDE.md. Put specialized workflows in skills. Put external service connections in MCP servers.
Summary Stats
| Category | Count |
|---|---|
| Anthropic Official Skills | 17 |
| Knowledge Work Plugin Skills | 47 |
| BMAD Method | 9 agents + 15 commands |
| Total | ~90+ available skills |