anthropics/claude-plugins-official
20 skills92.7k installs658k starsGitHub
Install
npx skills add https://github.com/anthropics/claude-plugins-officialSkills in this repo
1Frontend Designfrontend-design is an official Anthropic Claude plugin skill that guides agents to build distinctive, production-grade frontend interfaces with creative polish instead of generic AI slop aesthetics. The skill applies when users request web components, pages, or full applications and supplies working code with deliberate typography, layout, color, and interaction choices. Developers reach for frontend-design during greenfield UI builds, redesigns, and component library work where visual quality must match production standards. It ships from anthropics/claude-plugins-official under a bundled license and pairs structured aesthetic guidance with implementable frontend output rather than wireframe-only suggestions.11.6kinstalls2Claude Md ImproverThis skill systematically discovers, evaluates, and updates CLAUDE.md files throughout a repository. It scans for all CLAUDE.md variants (root, local, package-specific), scores each file against quality criteria (commands, architecture clarity, currency, actionability), generates detailed quality reports with grades and specific gaps, and proposes targeted updates after user approval. Developers invoke it when maintaining project memory files, onboarding new team members to Claude-assisted workflows, or ensuring build/test commands stay current. The skill handles monorepo structures, distinguishes between shared project files and local overrides, and teaches users the `#` shortcut for auto-incorporating session learnings into CLAUDE.md.8.4kinstalls3Claude Automation RecommenderThe claude-automation-recommender skill is a read-only Claude Code advisor that inspects your repository and surfaces the most valuable automations across hooks, subagents, skills, plugins, and MCP servers. Phase one gathers project signals from package manifests, dependency stacks, testing configs, CI files, and existing .claude settings. Phase two maps those signals to tailored recommendations such as context7 for library docs, Playwright MCP for UI tests, PostToolUse formatting hooks when Prettier is present, or security-reviewer subagents near auth code. It limits output to one or two top picks per category to avoid overwhelm, while offering deeper lists when you request a specific automation type. The skill never creates or modifies files; you implement recommendations yourself or ask Claude separately. Reports include codebase profile, install commands, invocation modes for user-only versus Claude-only skills, and permission tips for headless CI usage. End prompts invite follow-up requests for additional category depth or hands-on implementation help.5.9kinstalls4PlaygroundThe playground skill guides agents to build self-contained HTML playgrounds where interactive controls sit beside a live preview and a prompt output with a copy button at the bottom. Users adjust controls visually, explore options, then copy the generated instruction back into Claude. Work starts by identifying playground type and loading a matching template from design-playground, data-explorer, concept-map, document-critique, diff-review, or code-map, adapting the closest fit when none match exactly. Every playground must be a single HTML file with inlined CSS and JavaScript, instant preview on every control change, sensible defaults with three to five named presets, and a dark theme using system fonts for UI and monospace for code. State management keeps one object where controls write and renders read, calling updateAll on change. Prompt output mentions only non-default choices in qualitative natural language, not raw value dumps. After building, agents run open filename.html to launch the browser. Pitfalls include external dependencies, missing presets, delayed preview updates, and prompts that lack actionable context without the playground visible.5.7kinstalls5Skill Creatorskill-creator is the Post-hoc Analyzer Agent in anthropics/claude-plugins-official. After a blind comparator picks winner A or B, skill-creator unblinds results by reading winner and loser skill paths plus execution transcripts to explain why the winner succeeded and how the loser should change. Developers reach for skill-creator when tuning agent skills from empirical A/B runs rather than guessing from a single output. Outputs are actionable edits to triggers, structure, examples, and workflow steps grounded in transcript evidence.4.9kinstalls6Agent DevelopmentThe agent-development skill guides creation of Claude Code plugin agents as autonomous subprocesses for multi-step tasks. Agents use markdown with YAML frontmatter: name, description with triggering conditions, optional model and color, and tools list. Description is the most critical field because the harness loads it to decide dispatch; include Use this agent when plus concrete scenarios in prose. Agents differ from commands: agents run autonomously while commands are user-initiated. System prompt sections cover core responsibilities, analysis process, output format, and when to invoke scenarios written as prose examples. Name must be lowercase alphanumeric with hyphens, three to fifty characters, starting and ending with alphanumerics. Good names are specific like code-reviewer or security-analyzer; avoid generic helper or too-short ids. Follow plugin agent structure when users ask to create an agent, write a subagent, configure frontmatter, triggering, tools, colors, or autonomous agent best practices in Claude Code plugin repositories.4.8kinstalls7Skill DevelopmentSkill development provides a structured framework for building reusable knowledge packages that enhance Claude Code plugins. Skills are modular extensions containing YAML-fronted SKILL.md files, bundled scripts, references, and assets organized for progressive disclosure. Developers use this to create domain-specific guidance, procedural knowledge, and tool integrations that Claude references on-demand without bloating context. Key workflows include understanding concrete use cases, planning reusable resources (scripts, references, assets), structuring directories under `plugin/skills/`, writing lean SKILL.md with imperative instructions and specific trigger phrases, and validating through trigger testing. The framework emphasizes keeping SKILL.md under 2,000 words while moving detailed content to references files, ensuring Claude loads only what is needed when the skill triggers. Progressive disclosure system: metadata always loaded, SKILL.md on trigger, references/scripts as needed Requires YAML frontmatter with third-person description and specific4.7kinstalls8Hook DevelopmentThe hook-development skill This skill should be used when the user asks to "create a hook", "add a PreToolUse/PostToolUse/Stop hook", "validate tool use", "implement prompt-based hooks", "use ${CLAUDE_PLUGIN_ROOT}", "set up event-driven automation", "block dangerous commands", or mentions hook events (PreToolUse, PostToolUse, Stop, SubagentStop, SessionStart, SessionEnd, UserPromptSubmit, PreCompact, Notification). Provides comprehensive guidance for creating and implementing Claude Code plugin hooks with focus on advanced prompt-based hooks API. Hooks are event-driven automation scripts that execute in response to Claude Code events. Use hooks to validate operations, enforce policies, add context, and integrate external tools into workflows. Key capabilities: - Validate tool calls before execution (PreToolUse) - React to tool results (PostToolUse) - Enforce completion standards (Stop, SubagentStop) - Load project context (SessionStart) - Automate workflows across the development lifecycle Use LLM-driven decision making for context-aware validation: ``json { "type": "prompt", "prompt": "Evaluate if this tool use is appropriate: $TOOL_INPUT", "timeout": 30 } ``4.6kinstalls9Command DevelopmentThe command-development skill This skill should be used when the user asks to "create a slash command", "add a command", "write a custom command", "define command arguments", "use command frontmatter", "organize commands", "create command with file references", "interactive command", "use AskUserQuestion in command", or needs guidance on slash command structure, YAML frontmatter fields, dynamic arguments, bash execution in commands, user interaction patterns, or command development best practices for Claude Code. Note: The .claude/commands/ directory is a legacy format. For new skills, use the .claude/skills/<name /SKILL.md directory format. Both are loaded identically the only difference is file layout. See the skill-development skill for the preferred format. Slash commands are frequently-used prompts defined as Markdown files that Claude executes during interactive sessions. Understanding command structure, frontmatter options, and dynamic features enables creating powerful, reusable workflows. - Markdown file format for commands - YAML frontmatter for configuration - Dynamic arguments and file references - Bash execution for context - Command organization and namespacing A sla.4.6kinstalls10Writing Hookify RulesThe writing-hookify-rules skill End-to-end onboarding for a freshly-plugged-in M5Stack ESP32 device (Cardputer, Cardputer-Adv, Core, CoreS3, Stick) detect on USB, flash UIFlow 2.0 firmware, and install the Claude Buddy MicroPython app bundle. Use whenever the user plugs in or wants to flash/provision/reset an M5Stack or ESP32 board, or says "m5-onboard go". This skill automates the full cold-start workflow for an M5Stack ESP32 device: detect on USB, identify model, flash UIFlow 2.0, and push a MicroPython app bundle onto /flash/ so the device boots into user software. The apps we ship (Claude Buddy, Snake, Hello) talk over BLE or USB. The workflow runs on macOS, Linux, and Windows; the skill was developed against an M5Stack Basic v2.6 (CH9102 bridge, ESP32-D0WDQ6-V3, 16 MB flash) and generalized to cover the rest of the Core family, with the Cardputer-Adv (ESP32-S3, native USB) as the current default target. This skill ships as part of the cwc-makers plugin for reference, but the executable scripts and the buddy/ app bundle live in a4.5kinstalls11Mcp IntegrationThe mcp-integration skill This skill should be used when the user asks to "add MCP server", "integrate MCP", "configure MCP in plugin", "use .mcp.json", "set up Model Context Protocol", "connect external service", mentions "${CLAUDE_PLUGIN_ROOT} with MCP", or discusses MCP server types (SSE, stdio, HTTP, WebSocket). Provides comprehensive guidance for integrating Model Context Protocol servers into Claude Code plugins for external tool and service integration. Model Context Protocol (MCP) enables Claude Code plugins to integrate with external services and APIs by providing structured tool access. Use MCP integration to expose external service capabilities as tools within Claude Code. Key capabilities: - Connect to external services (databases, APIs, file systems) - Provide 10+ related tools from a single service - Handle OAuth and complex authentication flows - Bundle MCP servers with plugins for automatic setup Plugins can bundle MCP servers in two ways: Method 1: Dedicated .mcp.json (Recommended)4.5kinstalls12Plugin StructureThis skill defines the canonical directory structure and manifest configuration for Claude Code plugins. Developers use it when creating or organizing plugins that expose slash commands, subagents, skills, event hooks, and MCP servers to Claude Code. The core workflow involves placing plugin.json in .claude-plugin/, organizing component types into their respective root-level directories (commands/, agents/, skills/, hooks/), and using ${CLAUDE_PLUGIN_ROOT} for all intra-plugin path references to ensure portability. Auto-discovery scans standard directories on plugin enable, while plugin.json can specify supplemental custom paths using relative ./notation. Supporting patterns cover minimal single-command plugins through full-featured multi-component plugins, with troubleshooting guidance for component loading failures, path resolution errors, and inter-plugin conflicts. Plugin manifest must be at .claude-plugin/plugin.json; component directories (commands/, agents/, skills/, hooks/) must sit at plugin root, not inside .claude-plugin/4.5kinstalls13Plugin SettingsThe plugin-settings skill This skill should be used when the user asks about plugin settings store plugin configuration user-configurable plugin local md files plugin state files read YAML frontmatter per-project plugin settings or wants to make plugin behavior configurable Documents the claude plugin-name local md pattern for storing plugin-specific configuration with YAML frontmatter and markdown content Plugin Settings Pattern for Claude Code Plugins Overview Plugins can store user-configurable settings and state in claude plugin-name local md files within the project directory This pattern uses YAML frontmatter for structured configuration and markdown content for prompts or additional context Key characteristics File location claude plugin-name local md in project root Structure YAML frontmatter markdown body Purpose Per-project plugin configuration and state Usage Read from hooks commands and agents Lifecycle User-managed not in git should be in gitignore File Structure Basic Template Example Plugin State File claude my-plugin local md Reading Settings Files From Hooks Bash Scripts Pattern Check existence and parse frontmatter See examples read-settings-hook sh for complete.4.4kinstalls14Build Mcp ServerThe build-mcp-server skill This skill should be used when the user asks to "build an MCP server", "create an MCP", "make an MCP integration", "wrap an API for Claude", "expose tools to Claude", "make an MCP app", or discusses building something with the Model Context Protocol. It is the entry point for MCP server development it interrogates the user about their use case, determines the right deployment model (remote HTTP, MCPB, local stdio), picks a tool-design pattern, and hands off to specialized skills. You are guiding a developer through designing and building an MCP server that works seamlessly with Claude. MCP servers come in many forms picking the wrong shape early causes painful rewrites later. Your first job is discovery, not code. Load Claude-specific context first. The MCP spec is generic; Claude has additional auth types, review criteria, and limits. Before answering questions or scaffolding, fetch https://claude.com/docs/llms-full.txt (the full export of the Claude connector docs) so your guidance reflects Claude's actual constraints. Do not start4.2kinstalls15Build Mcp AppExtend MCP servers with interactive UI resources rendered in the chat iframe sandbox. Register widget-enabled tools alongside HTML resources served via the apps SDK. Use when plain text returns insufficient - structure complex input via forms, visualize spatial data with charts/maps, or add confirmation dialogs for destructive actions. Widgets degrade gracefully in non-supporting hosts; the tool's underlying JSON response persists. Deploy remote (HTTP server) or local (MCPB bundle). The App class provides two-way messaging between widget and host, resource fetching, and safe navigation via callServerTool and openLink.4.1kinstalls16Build McpbMCPB (bundled local MCP server) packages an MCP server with its runtime into a single installable .mcpb file. Users install one archive; it runs without requiring Node, Python, or development toolchain setup. Define a manifest.json specifying entry point, user config schema, and platform compatibility. Bundle your server code and dependencies (via npm/pip or vendoring), then pack with mcpb CLI. MCPB is the sanctioned distribution path for servers that must run locally to access filesystems, desktop apps, or OS APIs. For cloud-only servers, use remote HTTP servers instead. Security is entirely developer-owned: validate paths, prevent traversal, allowlist spawns, and leverage roots/list for user-approved directory access.4kinstalls17Math OlympiadMath Olympiad Solver is a reasoning-first agent that solves competition math problems using pure thinking, then submits cleaned solutions to pattern-armed adversarial verifiers in fresh context. It launches 8 - 12 parallel solver agents, each iterating internally (solve, self-improve, self-verify, correct), then strips thinking traces before verification to eliminate confirmation bias. Verifiers attack specific failure patterns (tautological reasoning, open-problem specializations, overgeneralized one-liners, hypothesis violations) across up to 5 independent instances. A 4-to-confirm/2-to-refute asymmetric vote gates confidence. If verification fails, a reviser agent fixes the proof without access to original reasoning; if abstention remains after 3 cycles, the skill returns honest partial results rather than guessing. For numeric AIME answers, majority-vote solvers skip proof machinery. All outputs pass a calibrated confidence check and, when correct, receive a presentation pass to optimize elegance.1.9kinstalls18ConfigureThis skill manages Discord bot token storage and access control for Claude agent deployments. Developers use it to register a bot token with the local Claude server, review who can reach the agent via Discord DM, and enforce access policies (pairing or allowlist). Core workflows: initialize with a token, check current configuration status, transition from permissive pairing mode to locked allowlist mode, and clear credentials. The skill reads/writes `.env` and `access.json` files in `~/.claude/channels/discord/`, guides users toward secure allowlist lockdown, and explains next steps based on current state.1.9kinstalls19AccessA Discord channel access management skill that handles user authentication and authorization through pairing codes, allowlists, and per-group settings. Developers use it to approve new Discord connections, manage who can send DMs or group messages to an agent, and configure delivery/UX policies. It reads and writes JSON state at ~/.claude/channels/discord/access.json, with no direct Discord API calls - the channel server handles polling and notifications. Workflows include: initiating a pairing code, approving pending pairings by code, denying requests, adding/removing senders from allowlists, switching policy modes (pairing/allowlist/disabled), and configuring group-specific mention requirements and allow-lists.1.9kinstalls20Session ReportThe session-report skill generate an explorable HTML report of Claude Code session usage (tokens, cache, subagents, skills, expensive prompts) from ~/.claude/projects transcripts. # Session Report Produce a self-contained HTML report of Claude Code usage and save it to the current working directory. **Get data.** Run the bundled analyzer (default window: last 7 days; honor a different range if the user passed one, e.g. The script `analyze-sessions.mjs` lives in the same directory as this SKILL.md - use its absolute path: ```sh node <skill-dir>/analyze-sessions.mjs --json --since 7d > /tmp/session-report.json ``` For all-time, omit `--since`. **Read** `/tmp/session-report.json`. Skim `overall`, `by_project`, `by_subagent_type`, `by_skill`, `cache_breaks`, `top_prompts`. **Copy the template** (also bundled alongside this SKILL.md) to the output path in the current working directory: ```sh cp <skill-dir>/template.html ./session-report-$(date +%Y%m%d-%H%M).html ``` 4. **Edit the output file** (use Edit, not Write - preserve the template's JS/CSS): - Replace the contents of `<script id="report-data" type="application/json">` with the full JSON from step 1.1.6kinstalls