Guide

Best Claude Code & agent Skills, by use case

The best agent Skill depends on the job. For code review, code-review-plugin; for security, trail-of-bits; for testing, webapp-testing; for DevOps, devops-engineer; for frontend, frontend-design; for SEO, the seo-geo-claude-skills bundle. Each is a verified, catalogued SKILL.md skill. Pick a task for the full ranked shortlist.

Every pick links to its skill page with the composite score, license, provenance, and per-agent install commands. We rank by fit for the common case, not by raw star count.

At a glance

The top pick for each job

JobTop pickWhy
Code Quality & Reviewcode-review-pluginPre-merge review on Claude Code where you want a consistent, quotable format with zero install.
Security & Auditingtrail-of-bitsPre-release vulnerability scans and ongoing security review on a codebase you control.
Testing & QAwebapp-testingFilling test gaps in a web app or smoke-testing a UI after a change.
DevOps & Infrastructuredevops-engineerGeneralist DevOps work where you want a strong default across clouds.
Frontend & UI Designfrontend-designGreenfield landing pages and marketing sites that should feel intentional, not templated.
SEO & Contentseo-geo-claude-skillsRunning SEO as a structured agent workflow rather than ad-hoc prompts.
Marketing & Growthcro-patternsOptimizing a landing or pricing page for conversion.

Go deeper

Ranked shortlist per job

Start here

New to agent Skills?

FAQ

Common questions

What are the best Claude Code skills?

The best skill depends on the job. For code review use code-review-plugin (Anthropic, bundled, /review). For security use trail-of-bits. For testing use webapp-testing. For DevOps use devops-engineer. For frontend use frontend-design (~277K installs). For SEO use the seo-geo-claude-skills bundle. For marketing, the right one is task-specific (cro-patterns, email-sequence-builder, and more). Each links to a ranked shortlist below.

Are these skills only for Claude Code?

No. SKILL.md is a cross-agent standard, and many of these run on Cursor, Codex, Gemini CLI, and Antigravity too — but real support varies per skill, so we list only verified-working agents on each skill page. frontend-design, for example, is verified across all seven agents we track.

How are these rankings decided?

Each skill gets a composite 0–100 score from nine inputs: install count, provenance, GitHub stars, recency, cross-agent compatibility, documentation, install ergonomics, license, and verification. On these guide pages we order by fit for the common case, not raw popularity. The methodology is public at /about/methodology.

What is the difference between a skill and an MCP server?

A Skill encodes methodology — how to do something — and loads into the current session. An MCP server gives the agent access to external systems (databases, APIs, browsers). Most real workflows use both. See skills vs subagents vs MCP vs plugins for the full distinction.

Browse all guides or jump to the full directory.