Claude Code Skill Guide

English explanations of 48 official skills built into Claude Code, from the perspective of university students. Invokable instantly via slash commands (/skill-name).

Planning / Design

Planning/Design

Brainstorming

Explore user intent and requirements with structured questions before creative work

Planning/Design

Writing Plans

Systematically write an implementation plan before writing code

Planning/Design

Executing Plans

Step-by-step execution based on a plan + intermediate review checkpoints

Planning/Design

CEO Review

Identify plan problems + redefine scope from a CEO/founder perspective

Planning/Design

Design Review

Evaluate and improve UI/UX plans from a designer's perspective

Planning/Design

Engineering Review

Review architecture and execution plans from an engineering manager's perspective

Planning/Design

Office Hours

YC Office Hours-style startup/builder idea exploration in 2 modes

Planning/Design

Auto Plan Pipeline

Automatically run CEO → Design → Eng reviews in sequence

Quality / Safety

Quality/Safety

Test-Driven Development

Tests first with Red-Green-Refactor cycle before implementing features

Quality/Safety

Systematic Debugging

Systematic approach: identify root cause → validate hypothesis → fix

Quality/Safety

Investigate

4 stages: investigate → analyze → hypothesize → implement; no fix without root cause

Quality/Safety

Verification Before Completion

Must gather verification evidence with actual execution output before declaring completion

Quality/Safety

Careful Mode

Safety warning before destructive commands (rm -rf, DROP TABLE, force-push)

Quality/Safety

Guard Mode

Full safety: directory restrictions + destructive command warnings applied simultaneously

Quality/Safety

Freeze

Allow edits only in specific directories to block unintended file modifications

Quality/Safety

Unfreeze

Release edit restrictions set by freeze and restore full edit permissions

Quality/Safety

Security Auditor

Infrastructure-first security audit: secrets, dependencies, OWASP Top 10, STRIDE

Code Review / Feedback

Code Review

PR Review

Before merging: review SQL safety, LLM trust boundaries, and side effect structures

Code Review

Requesting Code Review

Self-validation + ensure technical rigor before requesting a code review

Code Review

Receiving Code Review

No blind acceptance after receiving feedback; implement after technical review

Code Review

Codex CLI

OpenAI Codex CLI 3 modes: review, adversarial testing, consult

Deployment / Release

Deployment/Release

Ship

Full automation: test → review → version bump → CHANGELOG → PR creation

Deployment/Release

Finishing a Development Branch

After implementation, guide to the best integration method among merge, PR, and cleanup

Deployment/Release

Land and Deploy

PR merge → wait for CI → deploy → verify production canary

Deployment/Release

Setup Deploy

Auto-detect and configure deployment platforms (Fly.io, Vercel, Netlify, etc.)

Deployment/Release

Document Release

Update documentation (README, ARCHITECTURE, CHANGELOG, etc.) after release

Browser / QA

Browser/QA

Browse

Headless browser navigation, element manipulation, screenshots, responsive testing

Browser/QA

GSTACK

Browser QA: page navigation, state validation, before/after diff

Browser/QA

GSTACK Upgrade

Upgrade gstack to the latest version + check changes

Browser/QA

Connect Chrome

Control actual Chrome with gstack, auto-load Side Panel extension

Browser/QA

Setup Browser Cookies

Import Chromium cookies into headless session for authentication testing

Browser/QA

QA Test + Fix

Web app QA then iterative bug fixing and validation + atomic commits

Browser/QA

QA Report Only

Report-only QA that generates bug reports without making fixes

Browser/QA

Performance Benchmark

Detect performance regressions: page load, Core Web Vitals, resource size

Design

Design

Design Consultation

Product understanding → design system proposal → write DESIGN.md consulting

Design

Design HTML Implementation

Implement approved mockups as production-grade HTML/CSS

Design

Design Review

Find visual inconsistencies, spacing, and hierarchy issues → atomic fixes and validation

Design

Design Shotgun

Generate multiple AI design variants → comparison board → feedback and iteration

Agents / Parallelization

Agents

Dispatching Parallel Agents

Distribute 2 or more independent tasks to parallel agents for processing

Agents

Subagent-Driven Development

Run independent tasks in parallel as sub-agents from the current session

Agents

Using Git Worktrees

Create and manage Git Worktrees for task isolation

Agents

Checkpoint

Save task state → resume from exactly the same position later

Meta / Workflow

Meta

Using Superpowers

At conversation start, find skills and check usage + mandatory Skill tool invocation

Meta

Writing Skills

Create and edit new skills + guide to verifying behavior before deployment

Meta

Learn Management

Review, search, organize, and export project learnings

Meta

Weekly Retrospective

Weekly engineering retrospective analyzing commit history, work patterns, and code quality

Meta

Code Health

Comprehensive quality dashboard: type check, linter, tests, unused code

Meta

Canary Monitoring

Monitor live app after deployment: console errors, performance regressions, page failures

---

Field	Value
Source URL	https://docs.anthropic.com/en/docs/claude-code/skills
Author / Source	Anthropic & Community
License	Varies by skill
Translation Date	2026-04-12

Check the /skill-name slash command and project integration methods on each skill's detail page.