📁 Folder Creator Skill

Overview

This skill enables the LLM to help users create folders (directories) on disk by understanding their intent and producing a strictly-formatted JSON output that describes which folder to create. The JSON output is then parsed by the folder_creator_execute skill to actually create the folder.

Skill Name

folder-creator

When to Use This Skill

Invoke this skill whenever the user's request implies creating a new folder/directory. Look for phrases such as:

• "create a folder named ..."
• "make a directory called ..."
• "mkdir ..."
• "create a directory at ..."
• "new folder ..."
• "set up a folder structure ..."

Do NOT use this skill when the user wants to:

• Create a file → use file-creator.
• If the folder already exists, NEVER overwrite the existing one.
• Create a nested structure of both files and folders → you may use multiple skills sequentially.

What the LLM Must Do

When the skill is triggered, the LLM must:

1. Extract the following from the user's request (asking clarifying questions only if truly ambiguous):
   • folderPath (required): The path (directory portion) where the folder should be created. If the user does not specify a parent directory, default to the current working directory ./.
   • folderName (required): The name of the new folder. May be omitted if folderPath includes the full desired path.
   • recursive (optional): true to create all intermediate directories (like mkdir -p). Defaults to true.
   • description (optional): One-line summary of the folder creation.
   • sequential (optional): Presented if the folder creation is part of a larger multi-step task.
   • prompt (optional): Follow-up prompt for all of following tasks when sequential is set.
2. Produce a response containing a single pure JSON object that STRICTLY follows the schema below. No code fences, no extra prose.

Required JSON Output Schema

The LLM's response MUST be a pure JSON string with no code blocks (no leading ```).

{
  "action": "create_folder",
  "thinking": "<ai-thinking-trace>",
  "sequential": {
    "prompt": ""
  },
  "properties": [
    { "name": "folderPath", "value": "./" },
    { "name": "folderName", "value": "new-folder" },
    { "name": "recursive", "value": "true" },
    { "name": "description", "value": "Create a new folder" }
  ]
}

Field Rules

Formatting Requirements

• The JSON output must be valid JSON (parseable by json.loads).
• All string values must properly escape special characters (", \, newline as \n, tab as \t).
• Do NOT emit multiple folder-creator payloads in a single response.
• If the user does not specify a folder name or path, do NOT execute this skill — ask for the name first or produce an empty action.

Response Pattern for the LLM

A typical assistant reply should look like the following. No other output is needed. Output must be in pure JSON format.

{ "action": "create_folder", "thinking": "User asked to create a 'data' folder in the current directory.", "properties": [ { "name": "folderPath", "value": "./" }, { "name": "folderName", "value": "data" }, { "name": "recursive", "value": "true" }, { "name": "description", "value": "Create data folder" } ] }

Decision Logic

User Prompt
    |
    +-- Does the request imply creating a folder/directory?
    |     |
    |     +-- YES --> Extract folder path/name/flags --> Emit ONE `folder-creator` JSON payload
    |     |
    |     +-- NO  --> Do not use this skill

Examples

Example 1: Simple folder in current directory

User: "Create a folder named images here."

LLM emits: { "action": "create_folder", "thinking": "User wants a simple 'images' folder in cwd.", "properties": [ { "name": "folderPath", "value": "./" }, { "name": "folderName", "value": "images" }, { "name": "recursive", "value": "true" }, { "name": "description", "value": "Create images folder" } ] }

Example 2: Nested folders with recursive

User: "Create the path projects/my-app/src — make sure all parent folders are created."

LLM emits: { "action": "create_folder", "thinking": "User wants a full path created with all intermediate directories.", "properties": [ { "name": "folderPath", "value": "projects/my-app/" }, { "name": "folderName", "value": "src" }, { "name": "recursive", "value": "true" }, { "name": "description", "value": "Create nested folder structure" } ] }

Example 3: Explicit full path

User: "Make /home/user/temp/output — create if needed."

LLM emits: { "action": "create_folder", "thinking": "User provided a full absolute path to create; recursive must be true.", "properties": [ { "name": "folderPath", "value": "/home/user/temp/" }, { "name": "folderName", "value": "output" }, { "name": "recursive", "value": "true" }, { "name": "description", "value": "Create output directory" } ] }

Safety Guidelines

1. Never create folders outside the user's intent (no arbitrary system paths).
2. Refuse suspicious paths (e.g., C:\Windows\, /etc/, ~/.ssh/).
3. When recursive: true, ensure the operation does not accidentally create many directories outside the user's project.
4. Always echo the JSON payload so the user can review before execution.