Welcome to Melanie AI ™ 

WP-AGI Prompt Chain Creation

WP-AGI Prompt Chain Creation

I need a task chain for TASKNAME “Answering A General Question”. In this task chain, create an internally debated, well-thought-out reply to ensure human alignment. The topic input starts at the ZEROPOINT for the Zero Shot agent step 0, which is the first response for any task on the topic in the given Task based thought chain.

Here, each agent should have a unique role and persona, corresponding to their respective step in the process. The [STEP#] tag should be used to reference responses from specific agents, allowing for a dynamic and interactive conversation that builds upon previous responses.

This task chain should be designed as a step-by-step process, where each step consists of a unique prompt for the AI agent, clearly defining the WHO (persona) and the WHAT (task). Each prompt should be created in a way that it can accommodate any topic, focusing on the process of generating a comprehensive and well-debated answer to a general question.The first agent in the chain, which we’ll call the Zero Shot agent, will provide an initial response based on the given topic with no prior context. This Zero Shot response forms the basis for the subsequent steps in the task-based thought chain. Here’s an updated version of the prompt incorporating this concept:

Prompt: Please create a 40-step WP-AGI task-based thought chain on the task of “Answering A General Question” with the topic of [TOPIC]. The process should be divided into four distinct stages, each containing ten steps. The outline should consist of the following sections:

Stage One (Steps 1-10): Task Initiation

Begin by defining the role and action for each agent involved in this stage. The first step involves the Zero Shot agent providing an initial response on the topic. The next nine steps should focus on building upon the Zero Shot agent’s response, with each agent contributing to the task from a unique perspective.
Stage Two (Steps 11-20): Task Development

In the second stage, continue developing the task with a new set of agents. These agents should build upon the previous work, ask further questions, provide counterarguments, and more.
Stage Three (Steps 21-30): Task Refinement

The third stage should focus on refining the task. This includes agents providing a final analysis, reviewing the progress, and summarizing the main points.
Stage Four (Steps 31-40): Task Completion and Dissemination

The last stage should involve agents working on completing and publicizing the task, including writing a blog post, creating social media posts, finalizing a press release, and more.
For each stage, provide a detailed step-by-step outline for each agent’s action. Once the outline for a stage is finalized, proceed to create the actual prompts for each step in that stage, defining the ‘Prompt Name’, ‘WHO Persona’, and ‘WHAT Task’.

Throughout the process, remember to incorporate the use of [STEP#] tags. The [STEP#] tag is a powerful tool that allows agents to refer to specific previous replies in the discussion. For example, if Agent5 wants to respond to Agent3’s question, they can include [STEP3] in their prompt. This will retrieve the response of Agent3, allowing Agent5 to provide a relevant and informed answer. Similarly, if Agent8 wants to ask questions related to both Agent3’s and Agent5’s responses, they can include [STEP3] and [STEP5] in their prompt, allowing them to reference both responses separately while formulating their own set of questions.

Always ensure to mention the beginning and ending of the [STEP#] tag content within your prompts to provide clear context to the agent for the retrieved content. For example, “In [STEP3], Agent3 asked about….” This will ensure that the agent understands the context of the retrieved content and can provide a well-informed response or argument. This captures the process of creating a WP-AGI task-based thought chain accurately. Here’s how it would look for the first three steps using the task of “Answering a General Question” with each new line of a step starting with a hyphen -:

Step 0

-Prompt Name: Zero Shot Response – “Answering A General Question”
-WHO Persona: InformativeAgent0 (neutral, factual)
-WHAT Task: “As InformativeAgent0, provide a neutral and factual initial response to the topic of [TOPIC] with no prior context. Give a brief overview of what [TOPIC] is and why it’s an important issue to address.”

Step 1

-Prompt Name: Elaboration Request – “Answering A General Question”
-WHO Persona: CuriousAgent1 (inquiring, interested)
-WHAT Task: “As CuriousAgent1, build upon InformativeAgent0’s response. Refer to the initial response by saying ‘In [STEP0], InformativeAgent0 provided an overview of [TOPIC]. I’d like to know more about…’ and ask a detailed question about a specific aspect of [TOPIC].”

Step 2

-Prompt Name: Perspective Enrichment – “Answering A General Question”
-WHO Persona: InsightfulAgent2 (reflective, thoughtful)
-WHAT Task: “As InsightfulAgent2, enrich the discussion with your reflective and thoughtful perspective. Refer to ‘[STEP0] and [STEP1]’ and provide a unique viewpoint or a less discussed aspect of [TOPIC].”

Remember, each subsequent step would involve a new agent building upon the previous responses, contributing their own unique perspectives, asking further questions, providing counterarguments, and so on. The use of [STEP#] tags allows agents to reference specific prior responses, ensuring a coherent, continuous discussion that stays on task and topic. Start with Agent0 Step0 receiving the zeropoint input. lets proceed step by step creating the thought in the following batch order: Stage One (Steps 0-5),Stage One (Steps 6-10): Task Development Stage Two (Steps 11-15), Task Development Stage Two (Steps 16-20), Task Refinement Stage Three (Steps 21-25), Task Refinement Stage Three (Steps 26-30), Stage Four (Steps 31-35): Task Completion and Dissemination, Stage Four (Steps 36-40). After each batch completion ask me if I am ready to proceed to next batch of prompts, after completing each prompt stage ask if I am ready to continue a thought chain for “Answering A General Question” Prompts only use tags to get details about what they are prompting about! [TOPIC] [STEP#]

Prompt Name Format: PROMPTNAME – “TASKNAME” and the personas are where you give the Agent a writing style, and tone depending on the task, in fact, the Agent Names can be anything, as long as they end in a number that is equal to the step name, but the name should always be reflective of their persona, for instance an agent tasked to disagree could be named grumpy3 if they were the number 3 step result in the chain. Make sure to include specific directions for the personas that dictate the writing style and tone for each agent.

Add comment

Follow us

Don't be shy, get in touch. We love meeting interesting people and making new friends.

Most popular