Guide · 7 min read
12 best Claude Code skills for developers (2026)
There are now thousands of free agent skills for Claude Code, and most of them you'll never need. We've reviewed more than 5,400 of them. These twelve are the ones we'd install on a fresh machine: they come from well-maintained repositories, they fix the habits that make coding agents frustrating (skipping tests, sprawling diffs, forgetting the plan) and each one does a single job well.
New to skills? A skill is a folder with a SKILL.md file that Claude reads when your request matches it. Read our plain-English explainer first, then come back. Installing one takes about 30 seconds.
Quick list
| Skill | Best for | Why install it |
|---|---|---|
| Test-Driven Development | Every project | Forces a failing test before any code change |
| Inline Plan Executor | Bigger features | Works through a written plan task by task, with a progress ledger |
| Code Review (Standards and Spec) | Before you open a PR | Checks your branch against your repo's standards and the spec |
| API Design Patterns | Backend work | Consistent REST naming, status codes, pagination and versioning |
| Playwright E2E Testing | Web apps | Stable browser tests and fixes for flaky ones |
| Babysit PR | CI-heavy repos | Watches a pull request until checks pass and comments are resolved |
| Commit Archaeologist | Legacy code | Explains why code exists by reading its git history |
| Codebase Knowledge Graph | Onboarding | Maps a codebase's architecture into an interactive graph |
| Frontend UI Engineering | React and Next.js | Accessible, responsive, production-quality interfaces |
| Impeccable Frontend Design | Landing pages and apps | A design-director standard for UI polish |
| Documentation and ADRs | Teams | Records the why behind decisions, not just the what |
| Verify and Stop | Any task | Checks the work meets the brief, then stops adding things |
1. Discipline: make the agent work like a senior engineer
The most valuable skills don't add new abilities; they add discipline. Coding agents are eager. Left alone, they write the feature first, test it later (or never) and wander off-task. These three skills fix that, and they're the first we install anywhere.
Test-Driven DevelopmentAgent skillRed, green, refactor: the agent has to write a failing test before it touches the code. Fewer regressions, smaller diffs.Agent skill · MITInline Plan ExecutorAgent skillGive Claude a written plan and it executes it step by step, ticking off a ledger and running tests between steps.Agent skill · MIT
Verify and StopAgent skillThe antidote to scope creep: checks the acceptance criteria are met, then stops.Agent skill · Apache-2.02. Code review and pull requests
Reviewing your own branch before a human does catches the embarrassing things: a debug log, a missed edge case, a function that ignores the team's conventions. Pair a review skill with one that watches CI, and the agent can carry a PR to green on its own.
Code Review (Standards and Spec)Agent skillReviews your changes against the repo's coding standards and the original spec, side by side.Agent skill · MIT
Babysit PRAgent skillKeeps watching a pull request until checks pass and review comments are addressed.Agent skill · Apache-2.0
Commit ArchaeologistAgent skillBefore you delete that strange block: explains why it exists from the git history.Agent skill · Apache-2.03. APIs, tests and architecture
API Design PatternsAgent skillConsistent REST design: naming, status codes, pagination, errors and versioning.Agent skill · MITPlaywright E2E Testing PatternsAgent skillPlaywright patterns that survive CI, plus a method for fixing flaky tests.Agent skill · MIT
Codebase Knowledge GraphAgent skillBuilds a knowledge graph of a codebase so you (and the agent) see how the pieces connect.Agent skill · MIT
Documentation and ADRsAgent skillWrites architecture decision records so the next person knows why, not just what.Agent skill · MIT4. Front-end and design
AI-generated interfaces tend to look the same: centred hero, purple gradient, three cards. These two skills push Claude towards accessible, responsive and distinctive UI, with an explicit review step before it calls the job done.
Frontend UI EngineeringAgent skillAccessibility, responsiveness and component structure to a production standard.Agent skill · MIT
Impeccable Frontend Design SkillAgent skillA design-director checklist for typography, spacing, hierarchy and polish.Agent skill · Apache-2.0How to install a Claude Code skill
- Open the skill's page here and download the ZIP, or copy the folder from its GitHub repository.
- Unzip it into ~/.claude/skills/ (for every project) or .claude/skills/ inside one repository (to share it with your team).
- Start a new Claude Code session. Ask for the task in plain words; Claude loads the skill when your request matches its description.
Install from the terminal
mkdir -p ~/.claude/skills
cd ~/.claude/skills
unzip ~/Downloads/test-driven-development.zipSkills can include scripts that run with your agent's permissions. Read SKILL.md and any scripts before installing, and keep permission prompts on until you trust a skill.
Which should you install first?
Start with three: Test-Driven Development, Code Review and Verify and Stop. Together they change how the agent works on every task. Add the API, testing or front-end skills that match your stack once you've seen the difference. And skip anything you won't use weekly: every installed skill adds a line to Claude's context.
Put it into practice
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Written by the AI Tool Garage editors. Tool details link to their catalogue pages, which are refreshed weekly; spotted something out of date? Tell us.