anthropics/claude-code
10 skills130k installs1M starsGitHub
Install
npx skills add https://github.com/anthropics/claude-codeSkills in this repo
1Frontend Designfrontend-design is an agent skill from Anthropic’s Claude Code stack for solo and indie builders who need web interfaces that look intentional rather than templated. It fits the Build phase when you are turning requirements into components, marketing pages, or small product shells, and it also supports Validate when you need a credible prototype or landing surface before you commit to a full stack. The workflow starts with design thinking: clarify purpose and audience, pick a strong aesthetic tone (minimal, editorial, brutalist, retro-futuristic, and similar directions), respect technical constraints, and name what should be unforgettable. Only then does it implement working frontend code with attention to polish, hierarchy, motion where appropriate, and accessibility basics. Use it when chat-generated UI feels samey or when you want the agent to own visual direction as well as markup and styles. It is not a design-system package or a Figma replacement—it is procedural guidance for one-off builds inside your repo.41.4kinstalls2Agent DevelopmentAgent Development is an anthropics Claude Code catalog skill for solo and indie builders who need to add autonomous subagents to a plugin instead of bolting one-off prompts into chat. It walks through the markdown file format with YAML frontmatter, when to use agents versus slash commands, and how to write description blocks with concrete triggering examples so the host can route work reliably. You learn to shape system prompts with clear responsibilities, inherit or pin models, assign colors for UI distinction, and constrain tools like Read, Write, and Grep. The overview ties four concepts together: autonomous subprocess purpose, frontmatter contract, example-driven triggers, and prompt design. Use it whenever you are asked to create an agent, write subagent frontmatter, tune when-to-use descriptions, or follow agent development best practices inside Claude Code plugins.13.5kinstalls3Skill DevelopmentSkill Development is Anthropic’s meta skill for authoring Claude Code plugin skills. Indie builders and small teams invoke it when they want to create a new skill, add one to a plugin, sharpen a description for discovery, or reorganize long SKILL.md content using progressive disclosure. The material walks through what skills provide—specialized workflows, API and format integrations, domain playbooks, and repeatable bundled scripts—and the required folder layout with YAML frontmatter. It treats each skill as procedural knowledge models cannot fully internalize, which makes quality structure a force multiplier across every phase where you rely on agents. You still implement domain logic elsewhere; this skill keeps packaging, naming, and disclosure consistent so downstream sessions load the right capability without prompt bloat.13.1kinstalls4Mcp IntegrationMCP Integration is an agent skill for solo builders shipping Claude Code plugins who need external systems reachable as first-class tools. It walks through bundling MCP servers with plugins, choosing between a root `.mcp.json` and alternative configuration approaches, and wiring command args, environment secrets, and plugin-root-relative paths. The skill emphasizes practical integration patterns—stdio and remote transports, multi-tool surfaces from one service, and auth flows that go beyond API keys—so an agent can add databases, file systems, and SaaS APIs without ad-hoc scripting. Use it when prompts mention adding an MCP server, Model Context Protocol, or `${CLAUDE_PLUGIN_ROOT}` alongside MCP. It matters because MCP is the standard bridge between coding agents and production dependencies; getting config and auth right early avoids brittle one-off connectors later in ship and operate.10kinstalls5Command DevelopmentCommand Development is an Anthropic Claude Code skill for defining slash commands as Markdown prompts with optional YAML frontmatter. Solo and indie builders use it when they want repeatable agent workflows—code review rituals, release checklists, scaffold generators—without pasting the same long prompt every session. The skill covers command basics, frontmatter fields, dynamic arguments, file references, bash execution for context gathering, user interaction via AskUserQuestion, and how to organize or namespace commands across projects. It fits the Build phase on the agent-tooling shelf but also pays off later when you standardize how everyone on a small team invokes the same Claude behaviors. Pair it with other Claude Code customization skills when you are growing from ad-hoc chat into a small library of governed agent procedures.9.4kinstalls6Hook DevelopmentHook Development is an official Claude Code skill for building event-driven automation inside plugins. Solo builders who rely on agents for daily shipping eventually need guardrails: block destructive shell commands before they run, react to failed tool output, or inject repo context when a session starts. This skill walks through hook types with emphasis on prompt-based hooks that let the model reason about whether a tool call is appropriate, plus JSON configuration patterns and lifecycle coverage from UserPromptSubmit through Notification. It fits builders extending Claude Code rather than generic Cursor rules. Pair it with security-minded Ship practices when PreToolUse gates production changes. After hooks are in place, your agent workflow behaves more like a disciplined CI policy than a free-form chat session.9.1kinstalls7Plugin StructurePlugin Structure is an agent skill for solo and indie builders who ship custom Claude Code extensions instead of scattering slash commands and skills in ad hoc folders. It explains the conventional directory layout, what belongs in the manifest, and how each component type is organized and discovered at runtime. Use it when you are creating a new plugin, refactoring an existing bundle, or debugging why a command or skill is not loading. The skill emphasizes maintainable separation—manifest-driven configuration, predictable file naming, and root-relative paths—so your plugin behaves the same locally, in CI, and when teammates install it. It fits builders who treat the coding agent as part of the product surface and need a repeatable scaffold before adding business logic.8.9kinstalls8Plugin SettingsPlugin Settings is an agent skill for Claude Code plugin authors who need user-configurable behavior without hard-coding secrets or preferences. It documents the `.claude/plugin-name.local.md` pattern: YAML frontmatter for structured keys such as enabled toggles, scalars, and lists, plus a markdown section for task text, prompts fed back to Claude, or notes. Hooks, slash commands, and subagents read these files from the project directory so each repo can tune the same plugin differently. Solo builders shipping internal plugins or marketplace extensions use it when someone asks about plugin settings, storing configuration, reading frontmatter, or making behavior per-project. Files are user-managed, typically gitignored, which keeps machine-specific choices out of version control while still giving agents a single predictable path. The skill is meta-documentation of a convention—not a runtime loader—so implementation stays in your plugin code.8.8kinstalls9Writing Hookify RulesWriting Hookify Rules teaches your agent how to create and tune hookify configuration for Claude Code. Hookify rules are small markdown files with YAML frontmatter that match regex patterns on hook events—shell commands, file edits, stop attempts, or user prompts—and surface warnings or hard blocks. Solo builders use it when they want repeatable guardrails instead of one-off reminders in chat, such as flagging destructive rm patterns, blocking console.log in commits, or nudging review before the agent stops. The skill covers naming conventions, event selection, warn versus block actions, and message formatting. It fits anywhere you are hardening an agent workflow: while building features, before shipping risky refactors, or when operating with stricter change control. Invoke when the user explicitly asks to create, write, or configure a hookify rule or needs syntax for patterns and events.8.4kinstalls10Claude Opus 4 5 MigrationClaude Opus 4.5 Migration is a procedural agent skill for solo builders and small teams who already ship on Claude Sonnet 4.0, Sonnet 4.5, or Opus 4.1 and need a disciplined upgrade path without guessing platform-specific model strings. The skill walks the agent through searching the repository for model identifiers and API usage, replacing them with Opus 4.5 equivalents per hosting surface, removing the context-1m beta header where it would break, and applying the high effort setting that Opus 4.5 expects. It intentionally does not cover Haiku 4.5, so you avoid mixing migration scopes. After changes land, the agent summarizes diffs and invites prompt adjustments if behavior drifts—useful right before a release cut or when iterating on production agents after a provider rollout.7.2kinstalls