Best Dev Tools AI Skills — 31 Curated

Create, modify and A/B test custom Claude Skills with built-in performance tracking and an evaluation framework — your skills keep getting smarter.

Access Linear issues, projects and docs right inside Cursor — get task context without window switching and ship decisions faster.

Bring production monitoring into your dev flow: error logs, performance metrics and alerts next to your code for fast incident triage.

GitLab repos and CI/CD in one flow: commit code, watch pipeline status and manage merge requests from Cursor.

The official MCP server development guide: from an API to a working MCP server in Python or TypeScript, with tool-design best practices baked in.

The official Playwright testing skill: Claude drives your local web app like a real user — clicking, typing, screenshotting — to find and reproduce bugs.

14k+ stars: automatically convert any documentation site, GitHub repo or PDF into a Claude Skill, with conflict detection — your private knowledge becomes an AI skill.

A 19k-star one-stop collection: 337 skills + 30 agents + 70 commands spanning engineering, marketing, product, compliance, finance and C-level advisory.

Official skills from elite security firm Trail of Bits: static analysis, differential review, constant-time analysis, smart-contract audits and Semgrep rule writing.

28 official Sentry skills: auto-detect your stack and wire up the SDK, fix production errors with full context, review code against Sentry data.

OpenAI's official skill: when PR checks fail, skip the log-diving — it inspects GitHub Actions failures, finds the root cause and lands the fix.

Built by this site: a work-discipline protocol that makes Opus 4.8-class models deliver Fable-5-grade quality — plan gates, task cards, adversarial self-checks and real-product verification buy single-pass quality.

A complete software-development methodology for coding agents, built from composable skills. Your agent first pins down what you're really building, drafts a spec you approve in chunks, then runs a red/green TDD plan through subagents — all triggered automatically.

A skill/plugin that makes your AI coding agent answer in terse "caveman" style — filler gone, output tokens cut by ~65%, while code, commands and error text stay byte-for-byte exact and nothing technical is lost.

The agent skills Matt Pocock uses every day for real engineering, not vibe coding. Deliberately small, easy to adapt and composable, and they work with any model. The core idea: run a grilling/alignment pass before the agent starts, so it doesn't build the wrong thing.

A set of AI skills open-sourced by Khazix — the ones he actually uses daily. They follow the open Agent Skills standard and install on Claude Code, Codex, Cursor and 40+ other agents. Each is a structured instruction set battle-tested in real projects.

A GitHub-run community marketplace of Copilot customizations — hundreds of installable agents, instructions, skills, plugins, and workflows. Install one from the marketplace with a single command to give Copilot coding standards and task-specific abilities.

An AI-powered reverse-engineering and authorized-pentest skill router for coding clients (Claude Code, Codex CLI, Cursor, Cline). When an agent hits an APK, a binary, obfuscated JS, a CTF, or a pentest target, it routes to the right methodology, checks available tools, and runs a repeatable workflow instead of guessing commands. Covers 20+ security domains with auth/scope gating before any action.

A plugin and skills marketplace for Claude Code: 94 install-on-demand plugins bundling 203 domain agents, 175 skills, and 109 slash commands. Everything is generated from one Markdown source, so Codex CLI, Cursor, OpenCode, Gemini CLI, and GitHub Copilot consume it natively too.

Vercel's official collection of agent skills for AI coding agents, following the agentskills.io Agent Skills format. Drop it into Claude Code, Cursor or Codex and your agent audits React/Next.js performance, checks UI accessibility, and optimizes Vercel deployment cost and docs voice using Vercel Engineering's own rules.

A Claude Code skill bundle for bug hunting and external red-team work: 82 skills, 15 slash commands, 681 disclosed-report patterns curated across 24 core vulnerability classes, plus enterprise identity and infrastructure attack matrices. Skills load automatically by topic and stay in scope.

A Claude agent skill that turns cryptic terminal ASCII diagrams and pipe-and-dash tables into polished, browser-ready HTML pages or slide decks. Great for architecture diagrams, diff reviews, plan audits, data tables and project recaps.

A provider-neutral skill for Codex, Claude Code, and other agents that designs MVPs and audits or refactors agent harnesses. It covers tool permissions, loop budgets, context and memory, orchestration, observability, evals, caching, connectors, and launch gates.

Provides Claude Code with 29 scripts for iOS builds, testing, and simulator automation. It favors low-token semantic navigation through the accessibility tree and supports xcodebuild, app lifecycle management, accessibility audits, visual diffs, and diagnostic snapshots.

Turns a local web project or hosted page into a Dock-launchable macOS app. It inspects the project, selects WebKit or a compatible fallback, creates the launcher and icon, and verifies ports, launch, warm relaunch, quit, and cleanup.

A full lifecycle method for creating, improving, validating, reviewing, and packaging Agent Skills, with architecture selection, no-skill baselines, binary evidence evaluation, blind comparison, held-out gates, and bounded self-update.

Adds deterministic secret-leak checks at Claude Code and Codex tool boundaries. It blocks sensitive-file reads, credential-like writes, and shell exfiltration patterns, with gitleaks, Git hooks, and CI as defense in depth.

When a user explicitly asks to go beyond the pragmatic answer, it maps 3–6 advanced data structures, distributed algorithms, frameworks, patterns, or infrastructure choices and ranks them by calibrated complexity, learning value, and concrete payoff conditions.

Guides coding agents to manage reversible local changes with jj in an existing Jujutsu repository while Git remains the remote, PR, CI, and audit contract. It covers operation-log recovery, noninteractive splitting, parallel workspaces, and safe handoff.

A local-first configuration and skills suite for Claude Code, Codex, and Gemini that constrains coding agents with hard session budgets, dangerous-command and secret-write guards, plugin audits, a sandbox entry point, and bounded PR repair loops.

Four real skills and companion rules for coding agents: compress prose, reduce code through a YAGNI ladder, and recoverably trim oversized tool output on supported hosts.