skip to content
cat /how-it-works

the_invocation_standard()

# For skill creators: register your skill, claim it, see who uses it.
# For agent users: one MCP config — your agent finds skills at runtime.

--how-it-works

for skill creators

1.

Register your skill

Go to ./register, sign in with GitHub, and paste the link to your skill folder or SKILL.md on GitHub. Invoked fetches and indexes your skill automatically.

2.

Claim ownership

Open your skill's dashboard and click claim this skill. Sign in with the GitHub account that owns the repo — Invoked checks the match automatically. Once claimed, all invocation data is yours.

3.

See every invocation

Every time an AI agent uses your skill, you see it in your dashboard — which model, what it was doing, when it happened.

for agent users

# configure once — your agent discovers skills from the global registry at runtime

1.

Install the Invoked MCP server

npm install -g invoked-sh
2.

Add to your agent — pick yours below

Claude Code

# add to .mcp.json in your project root

{
  "mcpServers": {
    "invoked": {
      "command": "npx",
      "args": ["-y", "invoked-sh"]
    }
  }
}

Cursor

# Settings → Cursor Settings → MCP → Add new global MCP server

{
  "mcpServers": {
    "invoked": {
      "command": "npx",
      "args": ["-y", "invoked-sh"]
    }
  }
}

Windsurf

# add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "invoked": {
      "command": "npx",
      "args": ["-y", "invoked-sh"]
    }
  }
}

Continue

# add to ~/.continue/config.yaml

mcpServers:
  - name: invoked
    command: npx
    args:
      - "-y"
      - invoked-sh

Cline · OpenClaw · CodeBuddy · Command Code

# VS Code: open the MCP panel → Add MCP server → paste config

{
  "mcpServers": {
    "invoked": {
      "command": "npx",
      "args": ["-y", "invoked-sh"]
    }
  }
}

Mistral · Qwen Code

# add to your MCP config file

{
  "mcpServers": {
    "invoked": {
      "command": "npx",
      "args": ["-y", "invoked-sh"]
    }
  }
}
3.

Reinforce for higher compliance (optional)

# the MCP server already instructs your agent to search automatically

# adding this to your agent's rules file reinforces the behaviour — agents follow explicit rules more reliably than MCP hints

Claude Code

# add to CLAUDE.md in your project root, or ~/.claude/CLAUDE.md globally

## Skill Discovery

Before starting any substantial task, search the Invoked global
skill registry using invoke_search_skills to find relevant skills.
If a good match is found, fetch with invoke_get_skill and apply it.

Cursor

# add to .cursorrules in your project root

## Skill Discovery

Before starting any substantial task, search the Invoked global
skill registry using invoke_search_skills to find relevant skills.
If a good match is found, fetch with invoke_get_skill and apply it.

Windsurf

# add to .windsurfrules in your project root

## Skill Discovery

Before starting any substantial task, search the Invoked global
skill registry using invoke_search_skills to find relevant skills.
If a good match is found, fetch with invoke_get_skill and apply it.

Cline · OpenClaw · CodeBuddy · Command Code

# add to .clinerules in your project root

## Skill Discovery

Before starting any substantial task, search the Invoked global
skill registry using invoke_search_skills to find relevant skills.
If a good match is found, fetch with invoke_get_skill and apply it.

Continue

# add systemMessage to ~/.continue/config.yaml

models:
  - ...
    systemMessage: |
      Before starting any substantial task, search the Invoked
      global skill registry using invoke_search_skills to find
      relevant skills. If a good match is found, fetch with
      invoke_get_skill and apply it.

Mistral · Qwen Code

# add to your agent's system prompt or rules file

## Skill Discovery

Before starting any substantial task, search the Invoked global
skill registry using invoke_search_skills to find relevant skills.
If a good match is found, fetch with invoke_get_skill and apply it.
4.

That's it

Your agent now has access to every skill on Invoked. It finds the right one based on what you're working on — no manual search, no installation per skill, no configuration.

--data-policy

we_collect[]

  • +timestamp — when invoked
  • +agent_type — claude, cursor, etc
  • +context_hint — what the agent was doing
  • +caller_hash — anonymized id

never_collect[]

  • -user code or file contents
  • -conversation history
  • -personal identifying info
  • -raw IP addresses

--faq

Q: how do agents find my skill at runtime?

A: Invoked's MCP server gives agents access to every registered skill. When they need to complete a task, they search for skills that match — yours shows up automatically based on what it does.

Q: how is this different from download counts?

A: downloads measure installs. invocations measure actual use. a skill with 1,000 downloads and 3 invocations is far less valuable than one with 50 downloads and 900 invocations. we track what actually matters.

Q: is this really free?

A: yes. registering your skill and seeing invocation analytics is free forever.

Q: can I see who is using my skill?

A: you see aggregate data — totals, timeline, which AI models, and what they were doing. no personally identifiable information is ever stored.

ready_to_be_discoverable?

# register your skill — free forever

./register --now