How do I export the dataset?

Use /mcts:mcts-dataset export with optional filepath and format; default is json. The export creates a backup file.

How does import merge work?

Import merges new prompts and updates existing by ID by default. You can choose replace to clear all prompts and import fresh.

What happens on reset?

Reset deletes all custom prompts and restores the original defaults. It requires explicit confirmation; exporting beforehand is recommended.

mcts-dataset

npx machina-cli add skill NewJerseyStyle/plugin-mcts/mcts-dataset --openclaw

Files (1)

SKILL.md

3.2 KB

MCTS Prompt Dataset Manager

Manage your prompt dataset for MCTS operations.

Available Operations

1. LIST - View all prompts

/mcts:mcts-dataset list

Use MCP tool: mcts_dataset_list

2. GET - Retrieve a specific prompt

/mcts:mcts-dataset get <prompt_id>

Use MCP tool: mcts_dataset_get with prompt_id

3. CREATE - Add a new prompt

/mcts:mcts-dataset create

Use MCP tool: mcts_dataset_create with:

name: Prompt name/identifier
category: Category (research/planning/coding/general)
template: The prompt template text
description: What this prompt is for
variables: List of template variables (e.g., ["problem", "context"])

4. UPDATE - Modify an existing prompt

/mcts:mcts-dataset update <prompt_id>

Use MCP tool: mcts_dataset_update with:

prompt_id: ID of prompt to update
Fields to update (name, category, template, description, variables)

5. DELETE - Remove a prompt

/mcts:mcts-dataset delete <prompt_id>

Use MCP tool: mcts_dataset_delete with prompt_id

EXPORT - Backup your dataset

IMPORTANT: Always export before importing to avoid data loss!

/mcts:mcts-dataset export

Use MCP tool: mcts_dataset_export with optional:

filepath: Custom export path (default: mcts_prompts_backup_<timestamp>.json)
format: json or yaml (default: json)

This creates a backup file you can use to restore your prompts.

IMPORT - Load prompts from file

WARNING: Import will merge with existing prompts. Duplicates will be overwritten.

Please export your current dataset first for backup!

/mcts:mcts-dataset import <filepath>

Use MCP tool: mcts_dataset_import with:

filepath: Path to the import file
merge_strategy: "merge" (default) or "replace"
- merge: Add new prompts, update existing by ID
- replace: Clear all and import fresh

RESET - Restore default prompts

DANGER: This will DELETE all custom prompts and restore defaults!

To reset, you must type "yes" to confirm:

/mcts:mcts-dataset reset

When prompted, type exactly: yes

This will:

Delete ALL current prompts (custom and modified)
Restore the original default prompt set
This action CANNOT be undone!

Recommendation: Export your prompts before reset!

Operation: $ARGUMENTS

Execute the requested dataset operation using the appropriate MCP tools.

For CREATE/UPDATE operations:

Ask user for the required fields
Validate the input
Execute the operation
Confirm success

For DELETE/RESET operations:

Show what will be affected
Request explicit confirmation
For RESET: require typing "yes"
Execute only after confirmation

For EXPORT:

Generate the backup file
Report the file path
Remind user to keep this safe

For IMPORT:

FIRST remind user to export current data
Validate the import file exists
Preview what will be imported
Execute the import
Report results

Source

git clone https://github.com/NewJerseyStyle/plugin-mcts/blob/main/skills/mcts-dataset/SKILL.mdView on GitHub

Overview

The MCTS-dataset skill lets you view, edit, backup, and restore prompts used in MCTS workflows. It provides CRUD operations (list, get, create, update, delete), plus export, import, and a reset to defaults to keep templates organized and safe.

How This Skill Works

It exposes MCP-based commands to manage prompts: list, get, create, update, delete, export, import, and reset. Data can be merged on import, and reset restores the default prompts after explicit confirmation, enabling safe maintenance of your MCTS templates.

When to Use It

View all prompts across categories to understand current coverage
Retrieve a specific prompt by ID to review or edit
Add, modify, or remove a prompt's name, category, template, description, or variables
Back up prompts or restore from a file using export and import
Reset to the original defaults after confirming to discard all custom prompts

Quick Start

Step 1: List prompts to inspect the current dataset: /mcts:mcts-dataset list
Step 2: Create or update a prompt by providing name, category, template, description, and variables via /mcts:mcts-dataset create or update
Step 3: Back up or restore prompts with /mcts:mcts-dataset export and /mcts:mcts-dataset import <filepath>, or reset after confirmation

Best Practices

Always export before making large changes or resets
Use descriptive names and clear descriptions for each prompt
Validate required fields (name, category, template, description, variables) before create/update
Review import merge strategy to control duplicates and updates
Test prompt templates after changes to ensure variable usage is correct

Example Use Cases

Back up the entire MCTS prompts dataset before refactoring templates
Import a new set of planning prompts with merge to add and update existing entries
Update a prompt to fix variable names or improve guidance within the template
Reset to default prompts after a project concludes and exports are saved
Audit prompts by listing and filtering by category (research/planning/coding/general)

Frequently Asked Questions

Add this skill to your agents