agentic-resources

name	agentic-resources
description	Curated directory of online resources for discovering prompts, agents, skills, and MCP tools. Use when searching for reusable AI coding assets, evaluating skill/tool marketplaces, installing community skills, finding MCP servers, or recommending resources to users. Covers official vendor sources, community marketplaces, and CLI tooling with assessment criteria.

Agentic Resources — Discover Prompts, Agents, Skills & MCP Tools

Structured reference of the best online sources for finding reusable AI coding assets across the agentic ecosystem. Covers official vendor resources, community-maintained marketplaces, CLI tools, and MCP server registries — with assessment criteria and usage patterns.

---

When to Use This Skill

• Searching for existing skills, agents, or MCP tools before building custom ones
• Evaluating which marketplace or registry to query for a specific need
• Installing community skills into the workspace
• Recommending external resources to users exploring the agentic ecosystem
• Deciding between multiple skill/tool sources for the same capability
• Onboarding new team members to the agentic tooling landscape

---

Resource Directory

1. MCP Tool Registries

Resource	Type	Access	Description
MCP Official Registry	Official	API: registry.modelcontextprotocol.io	Canonical registry by Anthropic. 1000+ servers. Searchable via API. Reverse-DNS naming (io.github.*/...).
Smithery	Community	Web: smithery.ai	Hosted MCP server marketplace. Servers run remotely via server.smithery.ai — zero local install.
mcp.so	Community	Web: mcp.so	Aggregating directory with categories, search, and config snippets. No public API.
Glama MCP Directory	Community	Web: glama.ai/mcp/servers	Searchable catalog with install instructions for Claude Desktop, VS Code, etc.
Awesome MCP Servers	Community	GitHub: punkpeye/awesome-mcp-servers	Curated awesome-list. Categorized (databases, dev tools, web, etc.).

MCP registry MCP servers (search registries from within agents):

Server	Transport	Install
io.github.formulahendry/mcp-server-mcp-registry	stdio (npm)	npx mcp-server-mcp-registry
io.github.wei/mcp-registry-mcp-server	stdio (npm)	npx mcp-registry-mcp-server
ai.shawndurrani/mcp-registry	SSE (remote)	URL: https://mcp-registry.shawndurrani.ai/sse

2. Skills Marketplaces

Resource	Type	Size	Stars	Description
SkillsMP	Community	375+ MCP skills, 130+ prompt skills	—	Largest marketplace. Search by keyword, sort by stars. Supports Claude Code, Copilot, Cursor, Roo, OpenCode.
davila7/claude-code-templates	Community	634 skills	19.7k	Largest single collection. Agents, prompt engineering, LLM frameworks, memory systems, tool design.
sickn33/antigravity-awesome-skills	Community	Agent orchestration	7.8k	Agent improvement and evaluation skills.
NeoLabHQ/context-engineering-kit	Community	Context engineering	401	Prompt engineering, agent evaluation, context patterns.
openclaw/skills	Community	Multi-domain	723	Community-contributed. Prompt engineering, security, various domains.
mrgoonie/claudekit-skills	Community	MCP management	1.6k	MCP management + utility skills.

Top individual skills by stars:

Skill	Author	Stars	Purpose
senior-prompt-engineer	davila7	19.7k	Prompt patterns, LLM optimization, agent design
prompt-engineering	davila7	19.7k	Prompting strategies, debugging agent behavior
fastmcp-client-cli	jlowin	22.7k	Query and invoke tools on MCP servers
create-mcp-server	RooCodeInc	22.1k	Create MCP servers from scratch
copilot-sdk	github	20.5k	Build agentic apps with GitHub Copilot SDK
openai-knowledge	openai	18.8k	OpenAI API docs, Responses API, MCP integration

3. Prompt Libraries

Resource	Type	URL	Description
Anthropic Prompt Library	Official	docs.anthropic.com/en/prompt-library	Curated prompts for various tasks.
Anthropic Prompt Engineering Guide	Official	docs.anthropic.com/en/docs/build-with-claude/prompt-engineering	Claude-specific prompting techniques.
OpenAI Prompt Engineering	Official	platform.openai.com/docs/guides/prompt-engineering	GPT-family prompt best practices.
Google AI Prompting Guide	Official	ai.google.dev/gemini-api/docs/prompting-intro	Gemini prompting strategies.
LangSmith Hub	Community	smith.langchain.com/hub	Community-shared prompt templates for LangChain.

4. Agent Frameworks & Collections

Resource	Type	URL	Description
VS Code Agent Mode	Official	.github/agents/*.agent.md	Native agent definitions with YAML frontmatter + markdown.
GitHub Copilot Extensions	Official	github.com/marketplace?type=apps&copilot_app=true	Copilot-compatible extensions (agents-as-apps).
Claude Code	Official	docs.anthropic.com/en/docs/claude-code	CLI agent with .claude/ skills/commands system.
Cursor Rules	Community	cursor.directory	Community-shared .cursorrules files by framework/language.
RooCode	Community	github.com/RooCodeInc/Roo-Code (22.1k stars)	VS Code agent with built-in skill system.
OpenAI Agents SDK	Official	github.com/openai/openai-agents-python (18.8k stars)	Python multi-agent orchestration + MCP integration.

5. CLI Tools

Tool	Command	Description
skills CLI (Vercel)	npx skills	Install, search, update agent skills. Backend: skills.sh.
	npx skills find [query]	Search skills across registries.
	npx skills add <owner/repo>	Install skills from GitHub repos.
	npx skills list	List installed skills.
	npx skills check / update	Check for and apply updates.
FastMCP CLI	fastmcp list <server>	Query tools available on an MCP server.
	fastmcp call <server> <tool>	Invoke a tool on an MCP server.

6. Curated Aggregators

Resource	Type	URL
awesome-mcp-servers	Community	github.com/punkpeye/awesome-mcp-servers
awesome-claude-code	Community	github.com/anthropics/awesome-claude-code
context-engineering-kit	Community	github.com/NeoLabHQ/context-engineering-kit

---

Methodology

Phase 1: Source Selection

Given a search need, pick the right source:

Looking For	Primary Source	Secondary Source	CLI Shortcut
MCP server to wrap an API	MCP Official Registry	mcp.so, Smithery	mcp-registry MCP server
Reusable coding skill	SkillsMP marketplace	davila7/claude-code-templates	npx skills find <query>
Prompt template	Anthropic Prompt Library	SkillsMP (prompt-engineering)	—
Agent definition reference	cursor.directory	VS Code agent mode docs	—
Claude-specific skill	davila7/claude-code-templates	mrgoonie/claudekit-skills	npx skills add davila7/claude-code-templates

Phase 2: Evaluate Before Installing

Not every skill or tool is worth adding. Apply assessment criteria:

Factor	Install	Skip
Relevance	Directly matches a recurring task	Tangentially related
Quality	High stars, active maintenance, clear docs	No stars, abandoned, unclear
Overlap	No existing skill covers this	Already covered by workspace skills
Portability	Agent-agnostic, no hard dependencies	Tied to specific agent framework
Size	Focused, <200 lines	Bloated, kitchen-sink approach

Phase 3: Install and Verify

Skills from GitHub repos:

npx skills add <owner/repo>           # Interactive — pick skills to install
npx skills add <owner/repo> -y        # Auto-accept all
npx skills add <owner/repo> -g        # Install globally (user-level)

MCP servers:

// .vscode/mcp.json — add stdio servers
{
  "servers": {
    "<name>": {
      "type": "stdio",
      "command": "npx",
      "args": ["<package-name>"]
    }
  }
}

// Remote servers (no install)
{
  "servers": {
    "<name>": {
      "type": "http",        // VS Code uses "http" for streamable-http
      "url": "<server-url>"
    }
  }
}

Verify after install:

• Skills: check they appear in .github/skills/ or agent's skill references
• MCP servers: VS Code Output → MCP panel → confirm "Running" state

---

Known API Endpoints

Working APIs discovered during research that can be queried directly:

API	Endpoint	Returns
skills.sh	GET https://skills.sh/api/search?q={query}	{ skills: [{ id, skillId, name, installs, source }], count }
MCP Registry	GET https://registry.modelcontextprotocol.io/v0/servers?search={query}&limit=N	Server list with packages, transports, metadata
SkillsMP	Via skillsmp-mcp-server MCP tools	Search, AI search, get content, install, list repo skills

---

Anti-Patterns

• Installing everything — Skills bloat context. Install only what you'll use regularly.
• Ignoring overlap — Check workspace skills before adding external ones. Duplicates waste context budget.
• Trusting star count alone — High stars often reflect the repo's popularity, not the individual skill's quality. Read the actual SKILL.md.
• Using streamable-http in VS Code config — VS Code expects "type": "http", not "streamable-http". The registry uses different transport labels than VS Code's config schema.
• Assuming remote MCP servers are reliable — Community-hosted remote servers go down. Prefer stdio (local) for critical tools.

---

Output Format

When recommending resources to a user:

### Recommended Resources for {task}

| Resource | Why | Install |
|----------|-----|---------|
| {name} | {relevance to task} | {install command or URL} |