Source-backed learning hub for agent skills, AI agent workflows, MCP tools, coding agents, agent frameworks, agent evaluation, agent safety, deployment readiness, and rollout playbooks.
This repo is an independent educational/resource guide. It explains how skills fit across Codex, Claude Code, GitHub Copilot, OpenClaw, Hermes, Cursor, Gemini CLI, LangGraph, MCP, evaluation tools, deployment surfaces, and team adoption. The canonical catalog remains agentskillexchange/skills, and Agent Skill Exchange is the public site.
Star this repo if you want a practical map of agent skills, MCP, coding agents, agent frameworks, evaluation, safety, deployment, and team rollout patterns.
| I want to... | Start here |
|---|---|
| Understand agent skills quickly | Agent Skills 101 |
| Compare skills, tools, MCP, agents, and workflows | Skills vs Tools vs MCP |
| Browse the best public entry point | Awesome Agent Skills |
| Build, review, and evaluate a first skill | Build, Review, Evaluate Path |
| Compare Codex, Claude Code, Copilot, Cursor, Gemini, OpenClaw, LangGraph, and MCP | Framework Comparison |
| Find official docs, repos, and source-backed resources | Resource Index |
| Evaluate quality, safety, and rollout readiness | Checklists and Templates |
| Browse the canonical skill catalog | agentskillexchange/skills |
| Visitor | Read first | Why |
|---|---|---|
| New to agent skills | Agent Skills 101 | Get the concept before choosing tools or frameworks. |
| Building a first reusable workflow | Skill Author Starter Kit | Turn one source-backed workflow into a reviewable skill draft. |
| Comparing runtimes or frameworks | Framework Comparison | Match Codex, Copilot, Claude Code, Cursor, Gemini, OpenClaw, LangGraph, MCP, and SDK surfaces to the job. |
| Reviewing skill quality | First Skill Review Checklist | Check setup, permissions, safety, and verification evidence quickly. |
| Rolling out with a team | Team Evaluation Starter Kit | Plan a bounded pilot with owners, risks, and evidence. |
- What agent skills are and where they fit in the agent ecosystem.
- How skills differ from tools, MCP servers, runtimes, and workflows.
- How popular ecosystems approach agent frameworks, multi-agent workflows, RAG agents, agent orchestration, agent evaluation, MCP tools, coding agent skills, and agent safety.
- How agent ops connects observability, evals, LLM traces, approval workflows, workflow automation, AI workflow automation, model gateways, and rollout evidence.
- How agent safety review connects prompt injection, tool abuse, secrets, runtime guardrails, red teaming, policy, and governance.
- How to design source-backed skills with clear setup, permissions, and evidence.
- How to evaluate skill quality without relying on popularity claims.
- How teams can pilot agent workflows with sandboxing, approval gates, and rollout checks.
- Where to find official framework, MCP, safety, observability, and adoption resources.
This is a source-backed educational guide for agent skills and AI agent workflows. It provides learning paths, visual maps, examples, rollout playbooks, evaluation templates, source-labeling guidance, and a curated resource index. It is not a catalog mirror and does not try to track every published skill.
| Need | Start here |
|---|---|
| Find the right page quickly | Docs Index |
| Learn the agent skills ecosystem | Learning Paths |
| Follow a structured track | Curriculum |
| Understand terminology | Glossary |
| Design better skills | Best-Practices Cookbook |
| Use short review artifacts | Checklists |
| Compare Codex, Claude Code, GitHub Copilot, OpenClaw, Hermes, Cursor, Gemini CLI, LangChain, LangGraph, MCP, and OpenAI Agents SDK | Framework pages |
| Explore popular agent ecosystems | Ecosystems |
| Plan deployment and runtime hosting | Deployment |
| Design evaluation and benchmark evidence | Evaluation |
| Run skill-backed workflows safely | Agent Ops |
| Review security, policy, and governance | Security |
| Review source-backed resources | Resource Index |
| Evaluate skill quality and agent safety | Quality Checklist |
| Run a bounded rollout | Playbooks and Templates |
| Goal | Go here |
|---|---|
| Find the best starting point | Docs Index |
| Understand agent skills | Agent Skills 101 |
| Follow a role-based path | Curriculum |
| Learn the vocabulary | Glossary |
| Build a first skill | Skill Design Patterns |
| Compare frameworks | Framework Comparison |
| Explore agent ecosystems | Popular Agent Ecosystems |
| Plan runtime hosting | Deployment Overview |
| Design eval evidence | Evaluation Overview |
| Review agent ops evidence | Agent Ops Overview |
| Review agent safety | Agent Safety Review Guide |
| See workflow stacks | Showcase Workflow Stacks |
| Find official resources | Resource Index |
| Review skill quality | Skill Evaluation Basics |
| Build a skill | Skill Author Starter Kit and Cookbook |
| Run a team pilot | Team Evaluation Starter Kit |
| Browse the canonical catalog | Agent Skill Exchange |
- You want to understand where agent skills fit across AI agents, coding agents, MCP, Model Context Protocol, Codex, Claude Code, GitHub Copilot, OpenClaw, Hermes Agent, Cursor, Gemini CLI, LangChain, LangGraph, and OpenAI Agents SDK.
- You need source-backed framework resources, visual maps, examples, skill evaluation templates, agent safety checks, verification guidance, or rollout playbooks.
- You want representative examples without copying the full catalog.
- The full skill catalog.
- The installable skill source of truth.
- Official vendor claims unless the claim is source-backed.
- Catalog-wide generated skill dumps.
Agent skills are reusable instructions, workflows, and tool-usage patterns that help agents perform repeatable work. A good skill explains when to use a tool, how to set it up, what safety checks matter, and how to verify the result.
This repo helps developers answer four questions:
- Where do skills fit in the agent stack?
- Which labs, frameworks, and runtimes support skill-like workflows?
- How should teams evaluate, write, and verify skills?
- Where can I find source-backed examples without confusing them for the main ASE catalog?
It covers MCP and the Model Context Protocol, verification, skill evaluation, rollout playbooks, popular agent ecosystems, and practical adoption evidence for teams working with agentic development and operations.
| Repo | Purpose | Use it for |
|---|---|---|
| agent-skills-resources | Companion guide | Learning, diagrams, framework links, best practices, and curated examples |
| agentskillexchange/skills | Canonical catalog | Actual skill files, generated indexes, categories, verification, and installable entries |
| agentskillexchange.com | Live marketplace | Browsing, search, skill pages, industry collections, and creation workflows |
| Layer | Examples | What skills add |
|---|---|---|
| Model/provider | OpenAI, Anthropic, Google | Model-specific setup and constraints |
| Runtime | Codex, Claude Code, GitHub Copilot, OpenClaw, Hermes, Cursor, Gemini CLI | Repeatable agent workflows |
| Framework | LangGraph, OpenAI Agents SDK, ADK | Orchestration patterns and state |
| Ecosystem | CrewAI, AutoGen, Semantic Kernel, LlamaIndex, Pydantic AI, Haystack, Strands, Agno | Agent framework concepts and reusable workflow patterns |
| Deployment | Vercel, Cloudflare Workers, Fly.io, Modal, AWS Bedrock Agents, Azure AI Foundry, Google Vertex AI Agent Builder | Hosted runtime, sandbox, secrets, and rollout readiness |
| Evaluation | SWE-bench, GAIA, τ-bench, AgentBench, HELM, Inspect AI, OpenAI Evals, Ragas, DeepEval, Braintrust | Benchmarks, task rubrics, regression tests, and evidence bundles |
| Ops | LangSmith, Langfuse, Phoenix, Weave, OpenTelemetry, HumanLayer, Composio, n8n, Zapier, Vercel AI | Observability, evals, approvals, automation, gateways, and rollout evidence |
| Security | OWASP LLM Top 10, NIST AI RMF, Guardrails AI, Lakera, promptfoo, garak, PyRIT, LlamaFirewall | Prompt injection, red teaming, guardrails, secrets, policy, and governance |
| Protocol/tooling | MCP, CLIs, APIs, browser tools | Tool setup, permissions, and usage recipes |
| Verification | tests, scans, traces, approvals | Evidence that the workflow worked |
flowchart LR
User["User or team"] --> Runtime["Agent runtime"]
Runtime --> Framework["Agent framework or coding agent"]
Framework --> Skills["Skills layer"]
Skills --> Tools["Tools, APIs, CLIs, MCP servers"]
Skills --> Context["Memory and context"]
Skills --> Safety["Verification and guardrails"]
Skills --> Ops["Schedules, crons, deployments"]
Safety --> Evidence["Tests, scans, logs, citations"]
Tools --> Outcome["Repeatable workflow outcome"]
Evidence --> Outcome
| Area | Role in the ecosystem | Guide |
|---|---|---|
| Codex | Coding agent and terminal workflow runtime | Codex |
| Claude Code | Coding agent with project workflows, tools, and automation | Claude Code |
| GitHub Copilot | GitHub-native coding assistant, cloud agent, CLI, SDK, MCP, and skills surface | GitHub Copilot |
| OpenClaw | Agent runtime for providers, crons, skills, tools, and channels | OpenClaw |
| Hermes | Self-improving agent with skills, memory, and agent-managed workflows | Hermes |
| Cursor | IDE agent environment with context, skills, and background agents | Cursor |
| Gemini CLI | Open-source terminal agent from Google | Gemini |
| LangChain / LangGraph | Agent orchestration and stateful workflow framework | LangChain / LangGraph |
| MCP | Protocol for connecting agents to tools and context providers | MCP |
This repo also covers broader agent ecosystems that are not always skill-native but strongly shape reusable agent workflows: CrewAI, AutoGen, Semantic Kernel, LlamaIndex, Pydantic AI, Haystack, Strands Agents, and Agno. See the Coverage Matrix for how they relate to agents, tools, workflows, MCP, memory, evals, observability, guardrails, and approvals.
Deployment guidance covers runtime hosting, sandbox and container execution, secrets and environments, and deployment readiness. Ecosystem pages cover Vercel Platform, Cloudflare Workers, Fly.io, Modal, AWS Bedrock Agents, Azure AI Foundry, and Google Vertex AI Agent Builder.
Evaluation guidance covers benchmark landscape, eval design, regression testing, and evidence capture. Ecosystem pages cover SWE-bench, GAIA, τ-bench, AgentBench, HELM, MLCommons AILuminate, Inspect AI, OpenAI Evals, Ragas, DeepEval, and Braintrust.
For team adoption, this repo also covers the operational layer around skills: observability and evals, human approval workflows, workflow automation, model gateways and policy, and rollout evidence. Ecosystem pages cover LangSmith, Langfuse, Arize Phoenix, Weights & Biases Weave, OpenTelemetry GenAI, HumanLayer, Composio, n8n, Zapier, Vercel AI SDK, and Vercel AI Gateway.
Security guidance covers agent safety review, prompt injection and tool abuse, data and secrets handling, runtime guardrails, red teaming and evals, and policy and governance. Ecosystem pages cover OWASP LLM Top 10, NIST AI RMF, Guardrails AI, Lakera, promptfoo, garak, PyRIT, and LlamaFirewall.
Every resource in data/resources.json uses one of four labels:
Official: vendor or project-owned documentation or repository.Lab: research lab, model provider, or frontier-lab material.Community: useful third-party material that is not official.ASE: Agent Skill Exchange site, repo, data, or documentation.
When a claim is not source-backed, leave it out.
- data/resources.json: structured source list.
- data/ase-skill-mapping.json: representative ASE skill examples by framework and workflow area.
- docs-index.md: compact front door for all major docs.
- learning/: framework-neutral learning paths for agent skills.
- glossary.md: concise vocabulary for skill design and adoption.
- cookbook/: practical best-practice recipes and anti-patterns.
- checklists/: short fillable review artifacts for skills, resources, MCP tooling, team pilots, and permissions.
- ecosystems/: popular agent ecosystem coverage and comparison.
- deployment/: runtime hosting, sandboxing, secrets, provider matrix, and deployment readiness guidance.
- evaluation/: benchmark landscape, eval design, regression testing, and evidence capture guidance.
- ops/: agent operations guidance for observability, evals, approvals, workflow automation, gateways, and rollout evidence.
- security/: safety, policy, governance, prompt injection, secrets, guardrails, and red-team guidance.
- diagrams/: framework-neutral Mermaid diagrams.
- awesome-agent-skills.md: Awesome-style curated front door for agent skill resources.
- framework-comparison.md: practical comparison of major skill and agent workflow surfaces.
- showcase/: real-world workflow stacks that combine frameworks, skills, pilot notes, and verification evidence.
- generated/resource-index.md: generated resource index grouped by source type, framework, and tag.
- generated/ase-skill-mapping-index.md: generated representative ASE skill index grouped by workflow and framework.
- generated/nav-index.md: generated navigation index for learning, cookbook, ecosystem, deployment, evaluation, ops, security, framework, workflow, example, case-study, playbook, checklist, template, diagram, contributing, and maintenance pages.
- generated/template-index.md: generated index of fillable pilot and review templates.
- generated/repo-stats.md: generated repository data snapshot.
- workflows/: visual workflow guides that show how skills fit into practical SRE, security, data, content, and research work.
- case-studies/: practical scenarios that connect 2-4 existing ASE skills into reviewable workflows.
- playbooks/: adoption guides for teams evaluating skill-based workflows.
- starter-kits/: short paths for coding-agent users, MCP users, team evaluators, and skill authors.
- templates/: fillable worksheets for evaluation, risk review, security review, rollout readiness, and post-pilot review.
- adoption-matrix.md: lightweight comparison of starting workflows, risk levels, rollout paths, and expected evidence.
Use the repo as a small maintenance loop:
- Add or revise source-backed resources.
- Map only existing ASE skill slugs.
- Review examples against the quality checklist.
- Run validation and freshness checks.
- Keep source labels honest when ownership changes.
python3 scripts/validate-resources.py
python3 scripts/validate-links.py
python3 scripts/audit-freshness.py
python3 scripts/generate-resource-index.py
python3 scripts/generate-skill-mapping-index.py
python3 scripts/generate-nav-index.py
python3 scripts/generate-template-index.py
python3 scripts/generate-repo-stats.pyNear-term improvements:
- Keep official resource links fresh.
- Add more practical examples for team adoption and rollout evidence.
- Expand security, policy, and governance guidance for agent workflows.
- Add more completed evaluation examples.
- Improve framework, ecosystem, deployment, and evaluation comparison pages as official docs evolve.