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

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

2. Skills Marketplaces

Top individual skills by stars:

3. Prompt Libraries

4. Agent Frameworks & Collections

5. CLI Tools

6. Curated Aggregators

Methodology

Phase 1: Source Selection

Given a search need, pick the right source:

Phase 2: Evaluate Before Installing

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

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:

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} |