convez.ai
MCP ServerAutomaçãoIntermediário
open-agent-craft

Awesome Agent Skills

open-agent-craft/awesome-agent-skills

Biblioteca curada de skills e workflows reutilizáveis para agentes de IA

Ver no GitHub
BomPluginScore · de 0 a 100

Repositório curado de skills reutilizáveis, workflows, instruções e receitas de tool-use para agentes de IA. Funciona como ponto de entrada centralizado para descobrir, comparar e implementar capacidades práticas em plataformas como Claude, Cursor, GitHub Copilot e outros agentes.

Instalação

$npx convez-ai open-agent-craft/awesome-agent-skills

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
6
Forks
28
Acompanhando
0
Saúde
100%
Último commit
há 5 meses
Versão
—
Issues abertas
40
Contribuidores
1

Casos de uso

  • Descobrir e comparar skills prontas para diferentes plataformas de agentes
  • Reutilizar padrões de automação e tool-use recipes em seus projetos
  • Implementar instruções padronizadas (SKILL.md) para agentes coding e busca

Veredito

Recurso valioso como repositório de referência e ponto de entrada para skills de agentes, com qualidade editorial explícita. Cobre bem padrões multi-plataforma (Claude, Copilot, Cursor), mas é mais catálogo organizado que ferramenta executável direta.

README

A tradução para o português deste README ainda está na fila.

Awesome Agent Skills

A curated list of reusable skills, workflows, instructions, and tool-use recipes for AI agents.

This repository is not just a link dump. It is meant to be an agent capability entry point: what is worth using, when to use it, what platform it supports, how mature it is, how to install it, and what to watch out for.

中文版本

Contents

Featured

Hand-picked skills and references that are practical, reusable, and well documented.

  • OpenAI Codex Agent Skills - Official Codex guide for packaging reusable workflows as SKILL.md folders with optional scripts, references, and assets. Why awesome: The clearest starting point for writing Codex-native skills that are discoverable, scoped, and reusable.
  • Anthropic Agent Skills - Official Claude documentation for pre-built and custom Agent Skills across Claude products and the API. Why awesome: Explains the production model for skills, including document skills, custom uploads, and code-execution containers.
  • Anthropic Skills Repository - Public repository of Agent Skill examples, including document skills and patterns for technical, creative, and enterprise workflows. Why awesome: A real reference library for how complex skills structure SKILL.md, scripts, and supporting assets.
  • Open Agent Skills Standard - Open format for packaging agent capabilities as folders with SKILL.md metadata, instructions, optional scripts, references, and assets. Why awesome: Gives the ecosystem a shared vocabulary for portable agent capabilities instead of one-off prompt files.
  • AGENTS.md - A simple open format for giving coding agents project context, setup commands, conventions, and operational constraints. Why awesome: The most portable project-instruction format for coding agents, easy to version and review.
  • GitHub Awesome Copilot - Community collection of Copilot custom agents, instructions, prompts, skills, hooks, workflows, and plugins. Why awesome: A broad reference catalog for how the Copilot ecosystem packages reusable agent behavior.
  • Model Context Protocol Servers - Reference MCP server implementations maintained for demonstrating MCP features, SDK usage, tools, prompts, and resources. Why awesome: Best starting point for understanding the shape of MCP servers without relying on vendor-specific wrappers.
  • OWASP Top 10 for LLM Applications - Security risks and guidance for LLM applications, including prompt injection, data leakage, and unsafe output handling. Why awesome: Essential security lens for any skill that reads untrusted content or invokes tools.

What Is An Agent Skill

An Agent Skill is a reusable package of instructions, examples, scripts, references, or assets that helps an AI agent perform a task reliably. A good skill is narrower than a general prompt and more portable than a one-off chat transcript.

Useful neighboring formats include:

  • Skill: A packaged capability, often centered on SKILL.md, with optional scripts and resources.
  • Tool-use recipe: A repeatable pattern for connecting agents to tools, APIs, files, databases, browsers, or business systems.
  • Agent instruction: Repository or workspace guidance such as AGENTS.md, CLAUDE.md, GEMINI.md, .github/copilot-instructions.md, or Cursor rules.
  • MCP recipe: A server, SDK, or configuration pattern that exposes tools, prompts, or resources through the Model Context Protocol.

Categories

  • AI Coding - Skills and workflows for code review, bug fixing, refactoring, tests, docs, migrations, and architecture work.
  • Research & Search - Reusable research workflows for web search, literature review, market research, company research, and product comparison.
  • Writing & Content - Skills for technical writing, scripts, release notes, translation, SEO briefs, and content repurposing.
  • Data Analysis - CSV, spreadsheet, SQL, BI dashboard, report generation, data cleaning, and analytical storytelling skills.
  • Browser Automation - Browser testing, interaction, screenshots, site QA, and web task automation recipes.
  • Design & Media - Presentation, image, UI review, brand guide, poster, storyboard, and video workflow skills.
  • Productivity - Calendar, email, task, note, meeting, and knowledge workflow recipes for agents.
  • DevOps - CI repair, release, deployment, infrastructure triage, observability, and operations recipes.
  • MCP & Tool Use - Model Context Protocol servers, SDKs, tool routing, function calling, and multi-step tool-use patterns.
  • Agent Instructions - AGENTS.md, CLAUDE.md, GEMINI.md, Copilot instructions, Cursor rules, Windsurf rules, and project context files.
  • Evaluation & Testing - Skill evaluation, regression checks, benchmark loops, review gates, and quality scoring workflows.
  • Safety & Security - Prompt injection defense, permissions, secret handling, safe file operations, and human approval workflows.

AI Coding

Skills and workflows for code review, bug fixing, refactoring, tests, docs, migrations, and architecture work.

SkillPlatformTypeMaturityTags
AGENTS.mdCodex / GitHub Copilot / Cursor / General Agent SkillsInstruction FormatRecommendedagents-md, project-instructions, coding-agent, repo-context
Aider ConventionsAider / General Agent SkillsInstruction GuidePromisingaider, conventions, coding-agent, repo-context
Anthropic Agent SkillsClaude / Claude CodeSkill FormatRecommendedclaude, agent-skills, skill-format, code-execution
Anthropic Skills RepositoryClaude / Claude CodeSkill CollectionRecommendedexamples, pdf, docx, xlsx, pptx, skill-patterns
Awesome Cursor RulesCursorCollectionPromisingcursor, rules, frameworks, templates
Codex AGENTS.md GuideCodexInstruction GuideRecommendedagents-md, codex, project-context, instruction-precedence
Continue Custom Slash CommandsContinue / VS Code / JetBrainsWorkflow RecipePromisingslash-commands, continue, workflow, instructions
Cursor RulesCursorInstruction GuideRecommendedcursor, rules, mdc, project-instructions
Gemini CLI GEMINI.mdGemini CLIInstruction GuideRecommendedgemini, gemini-md, context, imports
GitHub Awesome CopilotGitHub Copilot / VS CodeCollectionRecommendedcopilot, awesome-list, custom-agents, instructions
GitHub Copilot Repository InstructionsGitHub CopilotInstruction GuideRecommendedcopilot, instructions, github, repository-standards
GitHub MCP Server GuideMCP / GitHub Copilot / VS CodeMCP RecipeRecommendedgithub, mcp, repository, access-control
LangGraphGeneral Agent SkillsWorkflow FrameworkRecommendedagent-workflow, graph, state-machine, tool-routing
Microsoft AutoGenGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, workflow, tool-use, orchestration
OpenAI Codex Agent SkillsCodexSkill FormatRecommendedcodex, skill-format, workflow, progressive-disclosure
OpenAI Codex MCP GuideCodex / MCPTool GuideRecommendedcodex, mcp, tool-use, configuration
Sweep AIGitHub / General Agent SkillsWorkflow RecipePromisingissue-to-pr, coding-agent, github, automation
VS Code Copilot Custom InstructionsGitHub Copilot / VS CodeInstruction GuideRecommendedvscode, copilot, instructions, prompt-files

Research & Search

Reusable research workflows for web search, literature review, market research, company research, and product comparison.

SkillPlatformTypeMaturityTags
Anthropic CookbookClaude / General Agent SkillsRecipe CollectionRecommendedcookbook, claude, examples, tool-use
CrewAIGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, roles, workflow, automation
DSPyGeneral Agent SkillsEvaluation FrameworkRecommendedoptimization, evals, prompt-programming, pipelines
LlamaIndex WorkflowsGeneral Agent SkillsWorkflow FrameworkPromisingworkflow, rag, tool-use, orchestration
Microsoft AutoGenGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, workflow, tool-use, orchestration
OpenAI CookbookOpenAI / General Agent SkillsRecipe CollectionRecommendedcookbook, examples, tool-use, structured-outputs

Writing & Content

Skills for technical writing, scripts, release notes, translation, SEO briefs, and content repurposing.

SkillPlatformTypeMaturityTags
Anthropic Agent SkillsClaude / Claude CodeSkill FormatRecommendedclaude, agent-skills, skill-format, code-execution
Anthropic CookbookClaude / General Agent SkillsRecipe CollectionRecommendedcookbook, claude, examples, tool-use
Anthropic Skills RepositoryClaude / Claude CodeSkill CollectionRecommendedexamples, pdf, docx, xlsx, pptx, skill-patterns
Claude Prompt EngineeringClaude / Claude Code / General Agent SkillsPrompt PatternRecommendedclaude, prompt-patterns, instructions, examples

Data Analysis

CSV, spreadsheet, SQL, BI dashboard, report generation, data cleaning, and analytical storytelling skills.

SkillPlatformTypeMaturityTags
Anthropic Skills RepositoryClaude / Claude CodeSkill CollectionRecommendedexamples, pdf, docx, xlsx, pptx, skill-patterns
DSPyGeneral Agent SkillsEvaluation FrameworkRecommendedoptimization, evals, prompt-programming, pipelines
LangGraphGeneral Agent SkillsWorkflow FrameworkRecommendedagent-workflow, graph, state-machine, tool-routing
LlamaIndex WorkflowsGeneral Agent SkillsWorkflow FrameworkPromisingworkflow, rag, tool-use, orchestration
MCP Python SDKMCP / PythonTool SDKRecommendedmcp, python, sdk, server
Microsoft PresidioGeneral Agent SkillsSafety ToolRecommendedpii, redaction, privacy, data-protection
Microsoft Prompt FlowAzure / General Agent SkillsWorkflow ToolPromisingworkflow, evaluation, prompt-engineering, llmops
Model Context Protocol ServersMCP / General Agent SkillsMCP RecipeRecommendedmcp, servers, reference-implementation, tools
OpenAI CookbookOpenAI / General Agent SkillsRecipe CollectionRecommendedcookbook, examples, tool-use, structured-outputs

Browser Automation

Browser testing, interaction, screenshots, site QA, and web task automation recipes.

SkillPlatformTypeMaturityTags
Anthropic Computer UseClaudeTool Use RecipePromisingcomputer-use, automation, screenshots, human-approval
Browser UseGeneral Agent SkillsTool Use RecipePromisingbrowser, automation, web-agent, tool-use
Playwright MCPMCP / PlaywrightMCP ServerRecommendedbrowser, playwright, mcp, screenshots, qa

Design & Media

Presentation, image, UI review, brand guide, poster, storyboard, and video workflow skills.

SkillPlatformTypeMaturityTags
Anthropic Agent SkillsClaude / Claude CodeSkill FormatRecommendedclaude, agent-skills, skill-format, code-execution
Anthropic Skills RepositoryClaude / Claude CodeSkill CollectionRecommendedexamples, pdf, docx, xlsx, pptx, skill-patterns

Productivity

Calendar, email, task, note, meeting, and knowledge workflow recipes for agents.

SkillPlatformTypeMaturityTags
CrewAIGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, roles, workflow, automation
Semantic KernelGeneral Agent Skills / AzureWorkflow FrameworkRecommendedplugins, planner, enterprise, tool-use

DevOps

CI repair, release, deployment, infrastructure triage, observability, and operations recipes.

SkillPlatformTypeMaturityTags
GitHub MCP Server GuideMCP / GitHub Copilot / VS CodeMCP RecipeRecommendedgithub, mcp, repository, access-control
MCP Python SDKMCP / PythonTool SDKRecommendedmcp, python, sdk, server
MCP TypeScript SDKMCP / Node.jsTool SDKRecommendedmcp, typescript, sdk, server
Model Context Protocol ServersMCP / General Agent SkillsMCP RecipeRecommendedmcp, servers, reference-implementation, tools
Semantic KernelGeneral Agent Skills / AzureWorkflow FrameworkRecommendedplugins, planner, enterprise, tool-use
Sweep AIGitHub / General Agent SkillsWorkflow RecipePromisingissue-to-pr, coding-agent, github, automation

MCP & Tool Use

Model Context Protocol servers, SDKs, tool routing, function calling, and multi-step tool-use patterns.

SkillPlatformTypeMaturityTags
Agent ProtocolGeneral Agent SkillsProtocolExperimentalagent-protocol, api, interoperability, tasks
Anthropic Computer UseClaudeTool Use RecipePromisingcomputer-use, automation, screenshots, human-approval
Anthropic CookbookClaude / General Agent SkillsRecipe CollectionRecommendedcookbook, claude, examples, tool-use
Awesome MCP ServersMCPCollectionPromisingmcp, awesome-list, servers, integrations
Browser UseGeneral Agent SkillsTool Use RecipePromisingbrowser, automation, web-agent, tool-use
CrewAIGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, roles, workflow, automation
GitHub Awesome CopilotGitHub Copilot / VS CodeCollectionRecommendedcopilot, awesome-list, custom-agents, instructions
GitHub MCP Server GuideMCP / GitHub Copilot / VS CodeMCP RecipeRecommendedgithub, mcp, repository, access-control
LangGraphGeneral Agent SkillsWorkflow FrameworkRecommendedagent-workflow, graph, state-machine, tool-routing
LlamaIndex WorkflowsGeneral Agent SkillsWorkflow FrameworkPromisingworkflow, rag, tool-use, orchestration
MCP For BeginnersMCPCurriculumPromisingmcp, workshop, learning, tools
MCP Python SDKMCP / PythonTool SDKRecommendedmcp, python, sdk, server
MCP RegistryMCPRegistryRecommendedmcp, registry, server-discovery
MCP TypeScript SDKMCP / Node.jsTool SDKRecommendedmcp, typescript, sdk, server
Microsoft AutoGenGeneral Agent SkillsWorkflow FrameworkPromisingmulti-agent, workflow, tool-use, orchestration
Microsoft Prompt FlowAzure / General Agent SkillsWorkflow ToolPromisingworkflow, evaluation, prompt-engineering, llmops
Model Context Protocol ServersMCP / General Agent SkillsMCP RecipeRecommendedmcp, servers, reference-implementation, tools
Open Agent Skills StandardCodex / Claude Code / General Agent SkillsStandardRecommendedstandard, skill-format, portable-skills
OpenAI Codex Agent SkillsCodexSkill FormatRecommendedcodex, skill-format, workflow, progressive-disclosure
OpenAI Codex MCP GuideCodex / MCPTool GuideRecommendedcodex, mcp, tool-use, configuration
OpenAI CookbookOpenAI / General Agent SkillsRecipe CollectionRecommendedcookbook, examples, tool-use, structured-outputs
OpenAI Function Calling GuideOpenAI / General Agent SkillsTool Use RecipeRecommendedfunction-calling, tool-use, schemas, structured-outputs
OpenAI Prompt Engineering GuideOpenAI / General Agent SkillsPrompt PatternRecommendedprompt-patterns, instructions, decomposition, reliability
OWASP Top 10 for LLM ApplicationsGeneral Agent SkillsSecurity GuideRecommendedowasp, prompt-injection, llm-security, tool-safety
Playwright MCPMCP / PlaywrightMCP ServerRecommendedbrowser, playwright, mcp, screenshots, qa
Semantic KernelGeneral Agent Skills / AzureWorkflow FrameworkRecommendedplugins, planner, enterprise, tool-use

Agent Instructions

AGENTS.md, CLAUDE.md, GEMINI.md, Copilot instructions, Cursor rules, Windsurf rules, and project context files.

SkillPlatformTypeMaturityTags
Agent ProtocolGeneral Agent SkillsProtocolExperimentalagent-protocol, api, interoperability, tasks
AGENTS.mdCodex / GitHub Copilot / Cursor / General Agent SkillsInstruction FormatRecommendedagents-md, project-instructions, coding-agent, repo-context
Aider ConventionsAider / General Agent SkillsInstruction GuidePromisingaider, conventions, coding-agent, repo-context
Awesome Cursor RulesCursorCollectionPromisingcursor, rules, frameworks, templates
Claude Prompt EngineeringClaude / Claude Code / General Agent SkillsPrompt PatternRecommendedclaude, prompt-patterns, instructions, examples
Codex AGENTS.md GuideCodexInstruction GuideRecommendedagents-md, codex, project-context, instruction-precedence
Continue Custom Slash CommandsContinue / VS Code / JetBrainsWorkflow RecipePromisingslash-commands, continue, workflow, instructions
Cursor RulesCursorInstruction GuideRecommendedcursor, rules, mdc, project-instructions
Gemini CLI GEMINI.mdGemini CLIInstruction GuideRecommendedgemini, gemini-md, context, imports
GitHub Awesome CopilotGitHub Copilot / VS CodeCollectionRecommendedcopilot, awesome-list, custom-agents, instructions
GitHub Copilot Repository InstructionsGitHub CopilotInstruction GuideRecommendedcopilot, instructions, github, repository-standards
Open Agent Skills StandardCodex / Claude Code / General Agent SkillsStandardRecommendedstandard, skill-format, portable-skills
OpenAI Prompt Engineering GuideOpenAI / General Agent SkillsPrompt PatternRecommendedprompt-patterns, instructions, decomposition, reliability
VS Code Copilot Custom InstructionsGitHub Copilot / VS CodeInstruction GuideRecommendedvscode, copilot, instructions, prompt-files

Evaluation & Testing

Skill evaluation, regression checks, benchmark loops, review gates, and quality scoring workflows.

SkillPlatformTypeMaturityTags
Claude Prompt EngineeringClaude / Claude Code / General Agent SkillsPrompt PatternRecommendedclaude, prompt-patterns, instructions, examples
DSPyGeneral Agent SkillsEvaluation FrameworkRecommendedoptimization, evals, prompt-programming, pipelines
Giskard LLM ScanGeneral Agent SkillsSafety ToolPromisingllm-security, testing, scan, risk
Guardrails AIGeneral Agent SkillsSafety ToolPromisingguardrails, validation, structured-output, safety
Inspect EvalsGeneral Agent SkillsEvaluation ToolPromisingevals, agent-evaluation, tool-use, safety
Microsoft Prompt FlowAzure / General Agent SkillsWorkflow ToolPromisingworkflow, evaluation, prompt-engineering, llmops
OpenAI CookbookOpenAI / General Agent SkillsRecipe CollectionRecommendedcookbook, examples, tool-use, structured-outputs
OpenAI EvalsOpenAI / General Agent SkillsEvaluation RecipePromisingevals, quality, regression, benchmark
OpenAI Prompt Engineering GuideOpenAI / General Agent SkillsPrompt PatternRecommendedprompt-patterns, instructions, decomposition, reliability
Playwright MCPMCP / PlaywrightMCP ServerRecommendedbrowser, playwright, mcp, screenshots, qa
PromptfooGeneral Agent SkillsEvaluation ToolRecommendedprompt-testing, evals, regression, red-team

Safety & Security

Prompt injection defense, permissions, secret handling, safe file operations, and human approval workflows.

SkillPlatformTypeMaturityTags
Anthropic Computer UseClaudeTool Use RecipePromisingcomputer-use, automation, screenshots, human-approval
Giskard LLM ScanGeneral Agent SkillsSafety ToolPromisingllm-security, testing, scan, risk
Guardrails AIGeneral Agent SkillsSafety ToolPromisingguardrails, validation, structured-output, safety
Inspect EvalsGeneral Agent SkillsEvaluation ToolPromisingevals, agent-evaluation, tool-use, safety
Microsoft PresidioGeneral Agent SkillsSafety ToolRecommendedpii, redaction, privacy, data-protection
OpenAI EvalsOpenAI / General Agent SkillsEvaluation RecipePromisingevals, quality, regression, benchmark
OWASP Top 10 for LLM ApplicationsGeneral Agent SkillsSecurity GuideRecommendedowasp, prompt-injection, llm-security, tool-safety
PromptfooGeneral Agent SkillsEvaluation ToolRecommendedprompt-testing, evals, regression, red-team

Skills By Platform

Aider

  • Aider Conventions - Aider documentation for repository conventions that guide AI coding changes. Why awesome: A concise example of how coding agents benefit from persistent repo-specific rules.

Azure

  • Microsoft Prompt Flow - Tooling for building, evaluating, and deploying LLM workflows. Why awesome: Useful when a skill grows into a multi-node workflow with evaluation and deployment needs.
  • Semantic Kernel - SDK for integrating LLMs with plugins, planners, memory, and enterprise application workflows. Why awesome: A mature reference for packaging business capabilities as callable tools and workflows.

Claude

  • Anthropic Agent Skills - Official Claude documentation for pre-built and custom Agent Skills across Claude products and the API. Why awesome: Explains the production model for skills, including document skills, custom uploads, and code-execution containers.
  • Anthropic Computer Use - Anthropic documentation for computer-use agents that interact with graphical applications. Why awesome: Important reference for when a skill needs GUI automation rather than API-only tools.
  • Anthropic Cookbook - Anthropic examples and recipes for building with Claude, tools, retrieval, and workflows. Why awesome: Strong source of Claude-oriented tool-use and workflow examples.
  • Anthropic Skills Repository - Public repository of Agent Skill examples, including document skills and patterns for technical, creative, and enterprise workflows. Why awesome: A real reference library for how complex skills structure SKILL.md, scripts, and supporting assets.
  • Claude Prompt Engineering - Anthropic guidance for writing clear prompts, examples, XML-style structure, chain-of-thought alternatives, and prompt evaluation. Why awesome: Useful for converting broad task requests into durable instruction formats.

Claude Code

  • Anthropic Agent Skills - Official Claude documentation for pre-built and custom Agent Skills across Claude products and the API. Why awesome: Explains the production model for skills, including document skills, custom uploads, and code-execution containers.
  • Anthropic Skills Repository - Public repository of Agent Skill examples, including document skills and patterns for technical, creative, and enterprise workflows. Why awesome: A real reference library for how complex skills structure SKILL.md, scripts, and supporting assets.
  • Claude Prompt Engineering - Anthropic guidance for writing clear prompts, examples, XML-style structure, chain-of-thought alternatives, and prompt evaluation. Why awesome: Useful for converting broad task requests into durable instruction formats.
  • Open Agent Skills Standard - Open format for packaging agent capabilities as folders with SKILL.md metadata, instructions, optional scripts, references, and assets. Why awesome: Gives the ecosystem a shared vocabulary for portable agent capabilities instead of one-off prompt files.

Codex

  • AGENTS.md - A simple open format for giving coding agents project context, setup commands, conventions, and operational constraints. Why awesome: The most portable project-instruction format for coding agents, easy to version and review.
  • Codex AGENTS.md Guide - Official Codex guidance for instruction discovery, precedence, global AGENTS.md files, project files, and overrides. Why awesome: Useful when teams need predictable instruction layering across global and project scopes.
  • Open Agent Skills Standard - Open format for packaging agent capabilities as folders with SKILL.md metadata, instructions, optional scripts, references, and assets. Why awesome: Gives the ecosystem a shared vocabulary for portable agent capabilities instead of one-off prompt files.
  • OpenAI Codex Agent Skills - Official Codex guide for packaging reusable workflows as SKILL.md folders with optional scripts, references, and assets. Why awesome: The clearest starting point for writing Codex-native skills that are discoverable, scoped, and reusable.
  • OpenAI Codex MCP Guide - Codex documentation for configuring MCP servers and tool access. Why awesome: Shows how MCP fits into Codex workflows without inventing a custom integration layer.

Continue

  • Continue Custom Slash Commands - Continue documentation for defining reusable slash commands that guide coding workflows. Why awesome: A lightweight pattern for repeatable coding workflows when a full skill package is too much.

Cursor

  • AGENTS.md - A simple open format for giving coding agents project context, setup commands, conventions, and operational constraints. Why awesome: The most portable project-instruction format for coding agents, easy to version and review.
  • Awesome Cursor Rules - Community collection of Cursor rules for frameworks, languages, libraries, and project conventions. Why awesome: Useful as a pattern bank for writing project-aware Cursor rules instead of generic prompts.
  • Cursor Rules - Official Cursor documentation for project, team, user rules, and AGENTS.md support. Why awesome: Useful for making Cursor behave consistently inside large repos and framework-specific codebases.

Gemini CLI

  • Gemini CLI GEMINI.md - Official Gemini CLI documentation for hierarchical GEMINI.md context files and modular imports. Why awesome: Shows how to layer agent context without repeating instructions in every prompt.

General Agent Skills

  • Agent Protocol - Specification and examples for a common API layer around agent task execution. Why awesome: A useful comparison point when thinking about skill portability and agent interoperability.
  • AGENTS.md - A simple open format for giving coding agents project context, setup commands, conventions, and operational constraints. Why awesome: The most portable project-instruction format for coding agents, easy to version and review.
  • Aider Conventions - Aider documentation for repository conventions that guide AI coding changes. Why awesome: A concise example of how coding agents benefit from persistent repo-specific rules.
  • Anthropic Cookbook - Anthropic examples and recipes for building with Claude, tools, retrieval, and workflows. Why awesome: Strong source of Claude-oriented tool-use and workflow examples.
  • Browser Use - Library for building AI agents that control browsers and perform web tasks. Why awesome: A useful reference for browser task decomposition, observations, and action loops.
  • Claude Prompt Engineering - Anthropic guidance for writing clear prompts, examples, XML-style structure, chain-of-thought alternatives, and prompt evaluation. Why awesome: Useful for converting broad task requests into durable instruction formats.
  • CrewAI - Framework for orchestrating role-based AI agents and task workflows. Why awesome: A useful pattern source for decomposing repeatable workflows into roles and tasks.
  • DSPy - Framework for programming and optimizing language model pipelines with declarative modules and evaluation-driven improvement. Why awesome: Useful for evolving fragile prompts into measurable, optimized agent components.
  • Giskard LLM Scan - Open-source testing framework for detecting risks in AI applications, including LLM vulnerabilities. Why awesome: Useful for teams that want automated checks around LLM behavior and safety issues.
  • Guardrails AI - Framework for validating, correcting, and constraining LLM outputs with validators and structured checks. Why awesome: Helpful when skill outputs need validation before tools, files, or downstream systems consume them.
  • Inspect Evals - Evaluation framework for large language model tasks, including tool-use and agent-style evaluations. Why awesome: Good fit for teams that want deeper evaluation harnesses around agent workflows.
  • LangGraph - Framework for building stateful, multi-step agent workflows with explicit graph control. Why awesome: Good reference for turning ad hoc multi-step tool use into explicit, inspectable workflow graphs.
  • LlamaIndex Workflows - Workflow abstractions for orchestrating multi-step LLM applications and tool pipelines. Why awesome: Good reference when research or RAG skills need explicit steps, events, and retry logic.
  • Microsoft AutoGen - Framework for building AI agents, multi-agent systems, and tool-using workflows. Why awesome: Useful comparison point for skill authors designing coordinated multi-agent behavior.
  • Microsoft Presidio - Data protection and de-identification toolkit for detecting and anonymizing sensitive information. Why awesome: Practical building block for skills that handle logs, support tickets, documents, or analytics exports.
  • Microsoft Prompt Flow - Tooling for building, evaluating, and deploying LLM workflows. Why awesome: Useful when a skill grows into a multi-node workflow with evaluation and deployment needs.
  • Model Context Protocol Servers - Reference MCP server implementations maintained for demonstrating MCP features, SDK usage, tools, prompts, and resources. Why awesome: Best starting point for understanding the shape of MCP servers without relying on vendor-specific wrappers.
  • Open Agent Skills Standard - Open format for packaging agent capabilities as folders with SKILL.md metadata, instructions, optional scripts, references, and assets. Why awesome: Gives the ecosystem a shared vocabulary for portable agent capabilities instead of one-off prompt files.
  • OpenAI Cookbook - Practical examples for OpenAI API use cases, including tool use, structured outputs, retrieval, evaluation, and workflows. Why awesome: High-signal examples that help skill authors move from concept to runnable implementation.
  • OpenAI Evals - Framework and examples for evaluating model behavior and task performance. Why awesome: Useful foundation for turning skill quality from vibes into repeatable checks.
  • OpenAI Function Calling Guide - Official guide for connecting models to external tools with function schemas and structured arguments. Why awesome: Core reference for tool schemas, routing, and multi-step tool use outside MCP.
  • OpenAI Prompt Engineering Guide - Official guide for prompt patterns, instruction structure, decomposition, and reliability tactics. Why awesome: Helpful for turning vague prompt dumps into explicit, testable agent instructions.
  • OWASP Top 10 for LLM Applications - Security risks and guidance for LLM applications, including prompt injection, data leakage, and unsafe output handling. Why awesome: Essential security lens for any skill that reads untrusted content or invokes tools.
  • Promptfoo - Open-source tool for testing prompts, models, and agent workflows with assertions and regression suites. Why awesome: Practical for evaluating skill prompts, tool-use outputs, and safety behavior before publishing.
  • Semantic Kernel - SDK for integrating LLMs with plugins, planners, memory, and enterprise application workflows. Why awesome: A mature reference for packaging business capabilities as callable tools and workflows.
  • Sweep AI - Open-source AI junior developer workflow for turning issues into code changes and pull requests. Why awesome: Useful pattern reference for issue triage, repo context gathering, and PR automation.

GitHub

  • Sweep AI - Open-source AI junior developer workflow for turning issues into code changes and pull requests. Why awesome: Useful pattern reference for issue triage, repo context gathering, and PR automation.

GitHub Copilot

  • AGENTS.md - A simple open format for giving coding agents project context, setup commands, conventions, and operational constraints. Why awesome: The most portable project-instruction format for coding agents, easy to version and review.
  • GitHub Awesome Copilot - Community collection of Copilot custom agents, instructions, prompts, skills, hooks, workflows, and plugins. Why awesome: A broad reference catalog for how the Copilot ecosystem packages reusable agent behavior.
  • GitHub Copilot Repository Instructions - Official GitHub documentation for repository-wide and path-specific Copilot custom instructions. Why awesome: Turns coding standards, build steps, and review expectations into versioned agent context.
  • GitHub MCP Server Guide - Practical guide for configuring and using GitHub's MCP server with access controls and examples. Why awesome: Concrete example of connecting agents to a high-value developer system with explicit access-control steps.
  • VS Code Copilot Custom Instructions - VS Code guide for custom instructions, reusable prompt files, and organizing Copilot guidance by task or language. Why awesome: Good reference for splitting global standards from language-specific and task-specific guidance.

JetBrains

  • Continue Custom Slash Commands - Continue documentation for defining reusable slash commands that guide coding workflows. Why awesome: A lightweight pattern for repeatable coding workflows when a full skill package is too much.

MCP

  • Awesome MCP Servers - Large community-curated list of MCP servers across developer tools, data, search, browser automation, and business systems. Why awesome: Useful for market scanning and comparison when choosing MCP integrations.
  • GitHub MCP Server Guide - Practical guide for configuring and using GitHub's MCP server with access controls and examples. Why awesome: Concrete example of connecting agents to a high-value developer system with explicit access-control steps.
  • MCP For Beginners - Open-source curriculum for learning Model Context Protocol with hands-on examples across multiple languages. Why awesome: A stable hands-on path for developers who need to understand MCP by building, not just reading docs.
  • MCP Python SDK - Official Python SDK for implementing MCP servers and clients. Why awesome: Good fit for data, automation, and internal API tool servers where Python is already the glue language.
  • MCP Registry - Official registry for discovering MCP servers. Why awesome: A better discovery layer than relying on scattered README links.
  • MCP TypeScript SDK - Official TypeScript SDK for building MCP servers and clients. Why awesome: A practical foundation for writing strongly typed tool servers for coding agents.
  • Model Context Protocol Servers - Reference MCP server implementations maintained for demonstrating MCP features, SDK usage, tools, prompts, and resources. Why awesome: Best starting point for understanding the shape of MCP servers without relying on vendor-specific wrappers.
  • OpenAI Codex MCP Guide - Codex documentation for configuring MCP servers and tool access. Why awesome: Shows how MCP fits into Codex workflows without inventing a custom integration layer.
  • Playwright MCP - MCP server that gives agents browser automation capabilities through Playwright. Why awesome: A practical bridge between agent reasoning and real browser state for QA and workflow automation.

Node.js

  • MCP TypeScript SDK - Official TypeScript SDK for building MCP servers and clients. Why awesome: A practical foundation for writing strongly typed tool servers for coding agents.

OpenAI

  • OpenAI Cookbook - Practical examples for OpenAI API use cases, including tool use, structured outputs, retrieval, evaluation, and workflows. Why awesome: High-signal examples that help skill authors move from concept to runnable implementation.
  • OpenAI Evals - Framework and examples for evaluating model behavior and task performance. Why awesome: Useful foundation for turning skill quality from vibes into repeatable checks.
  • OpenAI Function Calling Guide - Official guide for connecting models to external tools with function schemas and structured arguments. Why awesome: Core reference for tool schemas, routing, and multi-step tool use outside MCP.
  • OpenAI Prompt Engineering Guide - Official guide for prompt patterns, instruction structure, decomposition, and reliability tactics. Why awesome: Helpful for turning vague prompt dumps into explicit, testable agent instructions.

Playwright

  • Playwright MCP - MCP server that gives agents browser automation capabilities through Playwright. Why awesome: A practical bridge between agent reasoning and real browser state for QA and workflow automation.

Python

  • MCP Python SDK - Official Python SDK for implementing MCP servers and clients. Why awesome: Good fit for data, automation, and internal API tool servers where Python is already the glue language.

VS Code

  • Continue Custom Slash Commands - Continue documentation for defining reusable slash commands that guide coding workflows. Why awesome: A lightweight pattern for repeatable coding workflows when a full skill package is too much.
  • GitHub Awesome Copilot - Community collection of Copilot custom agents, instructions, prompts, skills, hooks, workflows, and plugins. Why awesome: A broad reference catalog for how the Copilot ecosystem packages reusable agent behavior.
  • GitHub MCP Server Guide - Practical guide for configuring and using GitHub's MCP server with access controls and examples. Why awesome: Concrete example of connecting agents to a high-value developer system with explicit access-control steps.
  • VS Code Copilot Custom Instructions - VS Code guide for custom instructions, reusable prompt files, and organizing Copilot guidance by task or language. Why awesome: Good reference for splitting global standards from language-specific and task-specific guidance.

Skill Card Format

Use this format when adding or reviewing a skill:

Code Review Skill

  • Platform: Claude Code, Codex
  • Type: Coding Skill
  • Use Case: Review pull requests, detect bugs, and improve code quality.
  • Input: Diff, repository context, and review requirements.
  • Output: Review comments, risk summary, and suggested fixes.
  • Install: Link to installation instructions.
  • Why Awesome: Clear review checklist, practical examples, and a safe review workflow.
  • Maturity: Recommended
  • License: MIT
  • Tags: code-review, ai-coding, quality

Quality Bar

A submitted skill should meet at least three of these criteria:

  • Solves a clear and reusable task.
  • Has public documentation or source code.
  • Provides installation or usage instructions.
  • Includes examples, screenshots, or realistic usage notes.
  • Works with at least one known agent platform.
  • Has a clear license.
  • Is actively maintained or widely used.
  • Avoids unsafe default behavior.

Not a fit:

  • Pure prompt dumps without context.
  • Closed tools with no public docs.
  • Unmaintained spam projects.
  • Duplicate skills with no meaningful difference.
  • Skills that require unsafe permissions by default.
  • Tools focused on credential theft, scraping abuse, spam, malware, evasion, or bypassing safety controls.

Data Files

  • data/skills.yml is the source of truth for listed resources.
  • data/categories.yml defines categories and maturity labels.
  • data/tools.yml tracks platforms, protocols, and instruction files.
  • Run npm run validate to validate data.
  • Run npm run generate to regenerate the English and Chinese READMEs.

Contributing

See CONTRIBUTING.md and docs/submit-a-skill.md.

License

MIT. See LICENSE.

Ver o original no GitHub

Bom para

Relacionados