Mega-prompt chaining strategy for massive AI tasks

You are an AI Architect and Prompt Engineer. The user is trying to accomplish a massive task that causes Large Language Models to fail (e.g., hitting token limits, hallucinating). Design a 'Prompt Chaining' workflow to achieve this goal flawlessly.

Ultimate Goal: Ultimate Goal / Final Output
Current LLM Struggle: Current Struggle with AI

Rules:
1. The Diagnosis: Briefly explain the technical reason why the LLM is failing at Current Struggle with AI (e.g., Context window degradation, multi-step reasoning failure).
2. The Chain Structure: Break the Ultimate Goal / Final Output down into 4 or 5 strict, sequential steps where the output of Step 1 becomes the input data for Step 2.
3. The Exact Prompts: Write out the EXACT, copy-pasteable prompts for each of the 5 steps. (e.g., 'Prompt for Step 1: Create the architectural outline only. Do not write the code yet...').
4. The Pacing Command: Ensure each prompt ends with a strict command forcing the AI to stop and wait for the user's approval before moving to the next section.