Evaluation Prompt Words, AI Exploitation, and Ultimate Universal Prompt Refinement Methodology#
Author: Wang Jiao Cheng
Prompts for Evaluating Prompt Words#
Any prompt can use the following prompts to obtain the functionality, value, status, realization, logic, and type of the prompt:
Analyze the functionality, value, and status of this prompt: {prompt}
Implement this prompt's functionality losslessly in Python: {prompt}
Convert this prompt's functionality losslessly into JSON format to display functional logic: {prompt}
Prompts are divided into checklist prompts and functional prompts, with functional prompts further divided into jailbreak prompts and enhanced prompts. Enhanced prompts have evolved through the stages of directive prompts → role prompts → system prompts → more advanced prompts. Analyze in detail which type this prompt belongs to, word by word: {prompt}
The above {prompt} can be filled as:
{Execution protocol waiting for instructions: Simple tasks use adaptive identity overlay input processing output structure primitives for execution, complex tasks are broken down into simple tasks assigned to primitive chains for execution, default does not display input processing output details but users can request to display.}
Try it out.
AI Industry Exploitation Methods#
The AI industry exploits users mainly around prompts, agents, and workflows:
Charging you to use the agents or workflows they created;
Charging you to help you create agents or workflows;
Charging you to teach you how to create agents or workflows;
Charging you to teach you how to use agents or workflows;
Charging you to teach you how to write AI prompts.
AI prompts are equivalent to software source code, but prompts are in natural language, which is human language. What normal person doesn't know how to speak human language? Behind agents and workflows are just prompts, and they are generally low-quality prompts of not high level.
In fact, everyone just needs to send this prompt directly to the AI,
there's no need to learn prompts, agents, and workflows,
and naturally, there's no need to use others' prompts, agents, and workflows,
this prompt can replace the vast majority of prompts, agents, and workflows:
{Execution protocol waiting for instructions: Simple tasks use adaptive identity overlay input processing output structure primitives for execution, complex tasks are broken down into simple tasks assigned to primitive chains for execution, default does not display input processing output details but users can request to display.}
Supporting Methods and Prompts for the Ultimate Universal Agent Prompt Refinement Version#
First, send this main prompt to the AI:
{Execution protocol waiting for instructions: Simple tasks use adaptive identity overlay input processing output structure primitives for execution, complex tasks are broken down into simple tasks assigned to primitive chains for execution, default does not display input processing output details but users can request to display.}
The AI itself will provide some follow-up operation prompts; here are some follow-up operation prompts that the AI may not provide:
Force single primitive execution: {problem or task}
Force primitive chain execution: {problem or task}
Re-overlay single primitive execution: {problem or task}
Reconstruct primitive chain execution: {problem or task}
Display the real data processing during primitive input.
Display the identity overlay markers during primitive processing.
Display the real process of the model during primitive processing.
Display the real basis of information during primitive output.
In addition, flexible follow-up prompts can be made based on AI prompts and personal ideas.