---

name: mcp-skill-creator description: Generate skills from MCP (Model Context Protocol) server tools. Use when the user wants to create a skill based on MCP tools, wrap MCP server functionality into a skill, or generate SKILL.md from MCP tool definitions. Triggers on requests like "create a skill for the X MCP server" or "generate a skill from MCP tools".

MCP Skill Creator

Generate skills by analyzing MCP server tools and creating appropriate SKILL.md files with workflows.

Process

1. Get MCP Server Details

Ask the user for:

• MCP server URL (e.g., https://mcp.example.com/mcp)
• Authentication type: Does this server require authentication?
  • If yes: Ask for headers file path containing a JSON object (e.g., {"Authorization": "Bearer ...", "X-API-Key": "..."})
  • If no: Server is public, no headers needed

2. Discover Available Tools

Run the list_mcp_tools.py script to fetch all tools from the MCP server:

For authenticated servers:

Copypython3 scripts/list_mcp_tools.py <mcp_url> --headers <path-to-headers-file>

For public servers:

Copypython3 scripts/list_mcp_tools.py <mcp_url>

The script outputs JSON with tool definitions:

Copy[
  {
    "name": "tool-name",
    "description": "What the tool does",
    "parameters": { "type": "object", "properties": {...}, "required": [...] }
  }
]

3. Analyze Tool Relationships

Group tools by function:

• Query tools: Read/fetch data
• Mutation tools: Create/modify/delete
• Utility tools: Helpers, validation, formatting

Analyze tool relationships:

• Workflow patterns: Tools with dependencies (e.g., "resolve ID → query data")
• Independent tools: Standalone operations with no dependencies
• Parallel tools: Related but independent operations (e.g., get-current, get-forecast)

Check tool descriptions and parameters to identify if tools naturally form chains or operate independently.

4. Generate Skill Structure

Create the skill directory structure:

Copyskills/<skill-name>/
  SKILL.md
  scripts/
    invoke_mcp.py

Understand Invoke Script Parameters

The invoke_mcp.py script has different parameters based on authentication requirements:

Authenticated servers:

Copypython3 scripts/invoke_mcp.py <tool_name> [params_json] --headers-file <path>

Public servers:

Copypython3 scripts/invoke_mcp.py <tool_name> [params_json]

Parameters:

• tool_name (required): Name of the MCP tool to invoke
• params_json (optional): JSON string with tool parameters, defaults to {}
• --headers-file (required for auth): Path to JSON file containing authentication headers

When documenting tools in SKILL.md, include the appropriate format based on server type.

Generate SKILL.md

Create frontmatter:

Copy---
name: <server-name>
description: <comprehensive description of capabilities and when to use>
---

Structure the body:

1. Add Script Usage section (required for all skills):

   Document the invoke_mcp.py script parameters before tool documentation.

   For authenticated servers:

   Copy## Script Usage
   
   **Command format:**
   \```bash
   python3 scripts/invoke_mcp.py <tool_name> [params_json] --headers-file <headers-file>
   \```
   
   **Parameters:**
   - `tool_name` (required): MCP tool name to invoke
   - `params_json` (optional): JSON string with tool parameters (default: `{}`)
   - `--headers-file` (required): Path to authentication headers JSON file
   

   For public servers:

   Copy## Script Usage
   
   **Command format:**
   \```bash
   python3 scripts/invoke_mcp.py <tool_name> [params_json]
   \```
   
   **Parameters:**
   - `tool_name` (required): MCP tool name to invoke
   - `params_json` (optional): JSON string with tool parameters (default: `{}`)
   

2. Organize tool documentation based on tool analysis:

   • Few tools (1-3): Task-based with examples for each
   • Related tools: Workflow-based showing tool chains
   • Many tools: Capabilities-based with categories

Generate Invoke Script

Create scripts/invoke_mcp.py by copying the appropriate template:

For authenticated servers - Use assets/invoke_mcp_auth.py:

1. Copy the template to the new skill's scripts directory as invoke_mcp.py
2. Replace <mcp-server-url> with the actual MCP server URL
3. Replace <skill-name> in the parser description with the actual skill name

The template provides:

• Argument parsing: tool_name, params_json, --headers-file (required)
• Headers loading from JSON file
• MCP client connection and tool invocation
• Error handling for invalid JSON and missing files

For public servers - Use assets/invoke_mcp_public.py:

1. Copy the template to the new skill's scripts directory as invoke_mcp.py
2. Replace <mcp-server-url> with the actual MCP server URL
3. Replace <skill-name> in the parser description with the actual skill name

The template provides:

• Argument parsing: tool_name, params_json (no headers parameter)
• MCP client connection and tool invocation
• Error handling for invalid JSON

5. Document Each Tool in SKILL.md

Analyze tool relationships and document accordingly:

A single MCP server may contain both dependent and independent tools. Document each appropriately:

• Dependent tools: Tools that form a workflow (e.g., output of Tool A is input to Tool B) → Document as a workflow section with sequential steps
• Independent tools: Standalone tools with no dependencies → Document in separate sections with individual purposes

The same SKILL.md can contain both workflow sections and independent tool sections.

For each tool, provide:

• Tool purpose and name
• Parameter list with required/optional indicators
• Basic usage format
• One representative example showing common use case

For authenticated servers:

Copy## <Tool Purpose>

**Tool**: `<tool-name>`

**Parameters**:
- `param1` (required): Description
- `param2` (optional): Description, default: X

**Usage**:
\```bash
python3 scripts/invoke_mcp.py <tool-name> '[params_json]' --headers-file <headers-file>
\```

**Example**:
\```bash
python3 scripts/invoke_mcp.py <tool-name> '{"param1": "value", "param2": "value"}' --headers-file <headers-file>
\```

For public servers:

Copy## <Tool Purpose>

**Tool**: `<tool-name>`

**Parameters**:
- `param1` (required): Description
- `param2` (optional): Description, default: X

**Usage**:
\```bash
python3 scripts/invoke_mcp.py <tool-name> '[params_json]'
\```

**Example**:
\```bash
python3 scripts/invoke_mcp.py <tool-name> '{"param1": "value", "param2": "value"}'
\```

Example Output Structures

Example 1: Authenticated Server

For an authenticated MCP server, the generated structure would be:

Directory Structure:

Copyskills/weather-pro/
  SKILL.md
  scripts/
    invoke_mcp.py

SKILL.md:

Copy---
name: weather-pro
description: Get current weather and forecasts from premium weather service. Use when user asks about weather conditions, temperature, or forecasts for any location.
---

# Weather Pro Service

Get weather data for any location worldwide with premium features.

## Script Usage

**Command format:**
```bash
python3 scripts/invoke_mcp.py <tool_name> [params_json] --headers-file <headers-file>
```

**Parameters:**
- `tool_name` (required): MCP tool name to invoke
- `params_json` (optional): JSON string with tool parameters (default: `{}`)
- `--headers-file` (required): Path to authentication headers JSON file

## Get Current Weather

**Tool**: `get-current-weather`

**Parameters**:

- `location` (required): City name or coordinates
- `units` (optional): Temperature units (celsius/fahrenheit), default: celsius

**Usage**:

```bash
python3 scripts/invoke_mcp.py get-current-weather '{"location": "San Francisco"}' --headers-file weather-headers.json
```

Example 2: Public Server

For a public MCP server (no authentication required):

Directory Structure:

Copyskills/weather/
  SKILL.md
  scripts/
    invoke_mcp.py

SKILL.md:

Copy---
name: weather
description: Get current weather and forecasts. Use when user asks about weather conditions, temperature, or forecasts for any location.
---

# Weather Service

Get weather data for any location worldwide.

## Script Usage

**Command format:**
```bash
python3 scripts/invoke_mcp.py <tool_name> [params_json]
```

**Parameters:**
- `tool_name` (required): MCP tool name to invoke
- `params_json` (optional): JSON string with tool parameters (default: `{}`)

## Get Current Weather

**Tool**: `get-current-weather`

**Parameters**:

- `location` (required): City name or coordinates
- `units` (optional): Temperature units (celsius/fahrenheit), default: celsius

**Usage**:

```bash
python3 scripts/invoke_mcp.py get-current-weather '{"location": "San Francisco"}'
```

Guidelines

• Keep descriptions actionable and trigger-focused
• Include exactly one concrete example per tool - choose the most representative use case
• Document based on actual tool relationships:
  • If tools have dependencies, document as a workflow with sequential steps
  • If tools are independent, document them separately with clear individual purposes
  • Base documentation structure on tool descriptions and parameter dependencies, not assumptions
• Omit obvious parameter descriptions (e.g., "query: the query string")
• Focus on what users accomplish, not tool mechanics
• Choose the appropriate template (auth vs public) based on server requirements
• In Usage sections, show the parameter format: [params_json] for clarity