skillsindex-mcp（skillsindex MCP 搜索）

SkillsIndex MCP

Search, score, and submit AI agent tools from SkillsIndex

What this skill does

Gives your AI agent read/write access to the SkillsIndex public directory — 11,000+ MCP servers, Claude skills, GPT actions, and IDE plugins, each scored on security, utility, maintenance, and uniqueness.

Security & transparency

• No credentials required — no API keys, no GitHub tokens, no secrets
• No local file access — the skill makes no filesystem operations
• Network calls only — all calls go to skillsindex.dev (public REST API over HTTPS) and npmjs.com (for npx to download the package)
• Read-only by default — search, get_tool, get_top_rated, check_security are all read-only
• Write operations — submit_tool POSTs a tool name + URL to skillsindex.dev/api/mcp/submit (public endpoint, no auth); subscribe POSTs an email to skillsindex.dev/api/subscribe
• Open source — full source code at https://github.com/thomasblc/skillsindex in the mcp/ directory
• No data exfiltration — the skill never reads or transmits your local files, environment variables, or system data

How it works

The skill runs as a local stdio MCP server via npx skillsindex-mcp. It bundles the compiled TypeScript and calls the SkillsIndex public API endpoints. No server is installed permanently — npx runs the package on demand.

Exact network endpoints called:

The Supabase anon key is a public read-only key — it only allows reading published tool data. This is the same key embedded in every SkillsIndex web page.

Tools

search_tools

Search by keyword, ecosystem, or category. Returns tools ranked by overall score.

query:      "postgres database MCP server"
ecosystem:  mcp_server | claude_skill | gpt_action | openclaw_skill | plugin
category:   databases-storage | browser-automation | code-execution | ...
limit:      1–20

get_tool

Full details for a tool by slug: security score, GitHub stats, pricing, install difficulty.

slug:  "filesystem-mcp-server"

get_top_rated

Top-scored tools, optionally filtered by ecosystem or category.

check_security

Security audit report: score (0–5), risk flags, audit notes based on static source analysis.

slug:  "tool-slug"     (preferred)
name:  "Tool Name"     (fuzzy match)

submit_tool

Submit a new tool for indexing. Sends name, URL, description to SkillsIndex review queue.

name:         "My MCP Server"
github_url:   "https://github.com/user/repo"
description:  "What the tool does"
ecosystem:    "mcp_server"
category:     "databases-storage"

subscribe

Subscribe an email to The Weekly Index — new tools + security alerts every Thursday.

Installation

No configuration needed. Add to claude_desktop_config.json:

{
  "mcpServers": {
    "skillsindex": {
      "command": "npx",
      "args": ["skillsindex-mcp"]
    }
  }
}

Data

• 11,000+ tools across 5 ecosystems
• Scoring formula: security (30%) + utility (30%) + maintenance (25%) + uniqueness (15%)
• Refreshed weekly via GitHub API
• Security scores from static analysis of open-source code (no proprietary data)

Source code

https://github.com/thomasblc/skillsindex → mcp/src/index.ts

Full source is auditable. The compiled output in dist/index.js matches the TypeScript source exactly.

Links

• Directory: skillsindex.dev
• npm: npmjs.com/package/skillsindex-mcp
• Source: github.com/thomasblc/skillsindex
• Scoring methodology: skillsindex.dev/methodology/