A curated collection of skills for Claude Code—reusable knowledge that Claude automatically applies when relevant.

Search and download ebooks from Anna's Archive

Search millions of books across formats (PDF, EPUB, MOBI, etc.) and download them using the fast download API.

Features:

• Search by title, author, or both
• Filter by format (PDF, EPUB, MOBI, AZW3, DJVU)
• Sort by year to get most recent editions
• Verify search results match expected title
• Download via fast API

Requirements:

• ANNAS_ARCHIVE_KEY environment variable (requires Anna's Archive membership)

Usage:

python3 annas.py search "Clean Code Robert Martin" --format pdf
python3 annas.py download <md5> --output ./books/

AssemblyAI Speech-to-Text and streaming transcription

Work with AssemblyAI's Speech-to-Text and LLM Gateway APIs, especially for streaming/live transcription, meeting notetakers, and voice agents.

Covers:

• Real-time streaming transcription
• Meeting notetaker integration
• Voice agent development
• LLM Gateway API usage
• Low-latency audio analysis

Write readable, maintainable code (based on Robert C. Martin)

Apply Clean Code principles when writing, reviewing, or refactoring code. Covers naming, functions, classes, error handling, and tests.

Key Principles:

• Names should reveal intent
• Functions should do one thing
• Classes should have single responsibility
• Handle errors explicitly, don't return null
• Tests should be fast, independent, and readable

Chapter References:

• chapters/names.md — Naming conventions
• chapters/functions.md — Function design
• chapters/classes.md — Class structure
• chapters/comments.md — When (not) to comment
• chapters/error-handling.md — Exception patterns
• chapters/objects-and-data.md — Data structures vs objects
• chapters/tests.md — Test design

Scrape documentation sites into markdown for AI context

Use slurp-ai to download and compile documentation websites into a single markdown file optimized for AI agent consumption.

Workflow:

1. Analyze sitemap first (node analyze-sitemap.js <url>)
2. Review page counts and section structure
3. Run slurp with informed --base-path and --max parameters

Features:

• Sitemap analyzer script for pre-scrape reconnaissance
• Guidance on scoping large documentation sites
• CLI reference for slurp-ai options

Requirements:

• Node.js v20+
• npm install -g slurp-ai

GoF and enterprise architecture patterns

Proven solutions to recurring software design problems. Use when designing architecture, refactoring, or when code shows symptoms like tight coupling or scattered responsibilities.

Pattern Categories:

Each pattern includes: intent, problem it solves, structure, implementation guidance, and real-world examples.

Extract text from EPUB, MOBI, and PDF files

Convert ebooks to plain text for analysis, processing, or reading. Pure Python extraction—no LLM calls.

Supported Formats:

Usage:

# Setup (one-time)
./setup.sh

# Extract
python3 scripts/extract.py input.epub output.txt
python3 scripts/extract.py input.pdf output.txt

Kalshi prediction market domain knowledge

Understand Kalshi's prediction markets—binary event contracts, pricing, order books, and API integration.

Covers:

• Binary contract mechanics (Yes/No outcomes, $0-$1 settlement)
• Price interpretation as implied probability
• Order book structure (bids, asks, spread)
• Data hierarchy: Series → Events → Markets
• REST API and WebSocket integration
• Authentication headers (KALSHI-ACCESS-KEY, etc.)

Read paywalled Medium articles

Fetch full content from paywalled Medium articles using free mirror services.

Services (in priority order):

1. Freedium — Best option, returns content directly
2. Archive.today — May require captcha
3. RemovePaywalls — Redirect page only

Usage:

Give Claude a Medium URL → skill automatically fetches via Freedium

Works with Medium-hosted publications: towardsdatascience.com, betterprogramming.pub, levelup.gitconnected.com, etc.

PostgreSQL 16 query construction and optimization

Comprehensive reference for PostgreSQL—query construction, optimization, schema management, and system introspection.

Covers:

• Standard and advanced SQL (CTEs, window functions, LATERAL joins)
• Query optimization and EXPLAIN analysis
• Index strategies (B-tree, GIN, GiST, BRIN)
• Schema management and migrations
• System catalog introspection
• JSON/JSONB operations
• Full-text search

Best Practices:

• Always use parameterized queries ($1, $2)
• Set statement timeouts for exploratory queries
• Use explicit transactions for multi-step operations

Root-cause debugging methodology

A disciplined approach to debugging that finds root causes before attempting fixes. Prevents random patches that mask underlying issues.

Core Process:

1. Reproduce the issue reliably
2. Form hypotheses about cause
3. Gather evidence systematically
4. Find root cause (not just symptoms)
5. Fix and verify

Includes:

• condition-based-waiting.md — Fix flaky async tests
• defense-in-depth.md — Layered error prevention
• root-cause-tracing.md — Systematic investigation
• find-polluter.sh — Find test pollution sources

Iron Law: No fixes without root cause investigation first.

Design modern JavaScript/TypeScript web applications

Comprehensive guidance for SPA/SSR/SSG/ISR/RSC architecture decisions, state management, performance, and team scaling.

Covers:

• Rendering strategy selection (CSR vs SSR vs SSG vs ISR vs RSC)
• State management patterns (local, global, server state)
• Component patterns (Container/Presentational, Compound, Hooks)
• Performance budgets and Core Web Vitals optimization
• Micro-frontends for large teams
• Security, accessibility, testing strategies

Reference Files:

• React patterns, SPA fundamentals, micro-frontends
• Performance, security, accessibility, testing
• Architecture decisions and trade-offs

Design effective web UI (forms, buttons, layouts, colors)

Apply interface design principles when building or reviewing web UI. Covers typography, forms, buttons, color systems, layouts, and accessibility.

Key Principles:

• Design exists to separate primary from secondary
• Users should instantly recognize what matters
• Good interface design is invisible

Reference Files:

• typography.md — Font sizes, line spacing, vertical rhythm
• forms-and-inputs.md — Validation, labels, touch targets
• buttons.md — Hierarchy, states, CTAs
• color-systems.md — Palettes, dark mode, tints/shades
• layout-and-spacing.md — Grids, spacing scales, responsive
• visual-hierarchy.md — Focus techniques, action pyramid
• accessibility.md — Contrast, focus states, WCAG

Sources: Web Interface Handbook, Refactoring UI, WCAG 2.1

Create effective Claude Code skills

Test-Driven Development applied to process documentation. Write skills that actually change agent behavior.

Process:

1. RED — Run scenario without skill, document baseline failures
2. GREEN — Write minimal skill addressing those failures
3. REFACTOR — Close loopholes found in testing

Covers:

• Skill structure and YAML frontmatter
• Claude Search Optimization (CSO) for discovery
• Testing skills with pressure scenarios
• Bulletproofing against rationalization
• Flowchart and code example conventions

Key Insight: If you didn't watch an agent fail without the skill, you don't know if the skill teaches the right thing.

git clone https://github.com/jaredsmith/claude-skills.git ~/claude-skills
ln -s ~/claude-skills/skills ~/.claude/skills

Update later with cd ~/claude-skills && git pull

If you want to add your own skills:

1. Fork this repo
2. Clone your fork: git clone https://github.com/YOUR_USERNAME/claude-skills.git ~/claude-skills
3. Symlink: ln -s ~/claude-skills/skills ~/.claude/skills

cp -r skills/clean-code ~/.claude/skills/

Skills are knowledge files that Claude Code automatically discovers and uses based on context. Unlike slash commands (which you invoke explicitly), skills are model-invoked—Claude reads the skill's description and activates it when relevant.

Each skill lives in a folder with a SKILL.md file containing structured knowledge, best practices, and procedures.

MIT