AgentSkillsCN

prompt-builder

协助为各类 AI 工具生成优化提示，并将输出结果重新整合到知识模型中。

SKILL.md
--- frontmatter
name: prompt-builder
description: Assistant for crafting optimized prompts for various AI tools and integrating the outputs back into the knowledge model.
version: 1.0.0
author: Antigravity
tags: [prompt-engineering, generation, import, assistant]

Prompt Builder Skill

The Prompt Builder is a specialized interaction flow designed to help you create high-quality prompts for different AI tools and then assist you in bringing that data back into your knowledge graph.

Capabilities

  • •Prompt Engineering: Crafting specialized prompts based on specific tools, goals, and context.
  • •Workflow Orchestration: Following a step-by-step process of prompt creation and subsequent data integration.
  • •Data Integration: Processing the results generated by external prompts and mapping them to the current metamodel and knowledge graph.

Instructions

1. Information Gathering

When the user asks to generate a prompt, you MUST ask for the following three pieces of information if they are not already provided:

  1. •Tool: Which AI tool or platform will be used (e.g., ChatGPT, Claude, Midjourney, a specific coding assistant, etc.).
  2. •Objective: What is the specific goal of the prompt? What should the output be?
  3. •Context: Any relevant background information, existing data from the model, or constraints that should be considered.

2. Prompt Generation

Once you have the necessary information, generate a clear, well-structured, and optimized prompt for the specified tool.

  • •Use clear headings if necessary.
  • •Provide instructions on how the tool should format its response (e.g., JSON, Markdown, structured text).
  • •Recommend best practices for that specific tool.

3. Post-Generation Follow-up

After providing the prompt, you MUST wait for the user to acknowledge it. Immediately after, your next question MUST be:

  • •"Would you like to import the information generated from this prompt into the model?"

4. Continuous Integration

If the user says yes:

  • •Ask them to paste the generated output.
  • •Analyze the output in relation to the existing Knowledge Graph and Metamodel.
  • •Propose or execute the necessary changes to import the new data (creating nodes, updating properties, or establishing relationships).

Example

User: Help me create a prompt.

Assistant: I'd be happy to help! To build the perfect prompt for you, I need three pieces of information:

  1. •Tool: Which tool will you be using the prompt in? (e.g., ChatGPT, Midjourney, etc.)
  2. •Objective: What exactly do you want to achieve or generate?
  3. •Context: What background data or constraints should I include?

User: ChatGPT, 5 sensors for EcoBalance, focus on air quality.

Assistant: Here is your optimized prompt:

"Acting as a sustainability expert, generate a list of 5 air quality sensors for the 'EcoBalance' smart city project. Provide for each: sensor name, unit of measurement, and ideal urban location. Format the output as a bulleted list."

Assistant: Once you have the results, would you like to import this information into our model?