Meta Prompter

What it does

The Meta Prompter skill optimizes prompts for AI models using a structured evaluation and refinement process. It takes an initial prompt as input, assesses its quality, clarifies ambiguities through questioning (potentially involving user interaction), builds an improved prompt based on the evaluation and clarifications, and then re-evaluates this new prompt. This iterative cycle continues until a satisfactory score is achieved or a maximum of three iterations are completed.

When to use it

• When you want to improve the quality and effectiveness of prompts for AI models.
• To overcome writer's block or generate more creative and nuanced outputs from an AI model.
• When initial prompt results are unsatisfactory, and you need a systematic way to refine them.
• For complex tasks where clarifying assumptions and context is crucial for achieving the desired outcome.

Key capabilities

• Prompt Evaluation: Assesses the quality of prompts using npx meta-prompter-mcp.
• Clarification Questioning: Identifies missing information or ambiguities in a prompt and generates questions to resolve them, potentially leveraging user input via the AskUserQuestion tool.
• Iterative Prompt Building: Constructs improved prompts based on evaluation results and answers to clarification questions.
• Re-evaluation & Gating: Repeatedly evaluates prompts and stops when a minimum score is reached or after three iterations.
• Output Saving (Optional): Saves the optimized prompt to a specified output directory using the Write tool.

Example prompts

• "Optimize this prompt: 'Summarize the key findings of this research paper.'"
• "Improve this prompt and save it to my assets folder: 'Translate this sentence into French.'"
• "Refine this prompt for a more creative response: 'Write a short story about a robot learning to love.'"

Tips & gotchas

• Session Management: When using the skill repeatedly, create and use a SESSION_ID to maintain consistency across files.
• Model Configuration: The --model flag is used for model resolution; refer to references/rules.md for details on configuring models.
• Iteration Limit: The process has a maximum of three iterations. If the prompt score remains below a threshold after that, the best-scoring prompt will be returned with a note about the final score.