📘 Overview

This file explains how to use the AI implementation planning prompt located in the same directory.
The prompt is designed to guide an advanced AI coding agent to generate a step-by-step technical plan for delivering a feature or fixing a ticket — without writing code yet.

🧠 What This Prompt Does

It instructs the AI to:

  • Ask for context about the feature/ticket
  • Collaborate using simulated domain experts (e.g., frontend, UX, architect)
  • Analyze the project’s codebase and propose changes
  • Collaborate with the human through Q&A and iteration
  • Generate a detailed implementation plan in a file named:
    plans/PLAN_<ISSUE_CODE>.md

🧑‍💻 How to Use It

✅ 1. Provide Context

Start by giving the AI the issue context. This can be:

  • Run this from the coding agent with the indexed codebase;
  • A Jira ticket code
  • A user story or acceptance criteria
  • Figma mocks, screenshots, diagrams, or requirements documents
  • External sources of information that can be accessed by the existing MCPs. Examples are: Confluence, Jira, Figma, Database, Github, Web Search.

💡 If you’re using a coding agent, simply paste the info or drag in files. The AI will take it from there.

🗣️ 2. Let the AI Ask Questions

The AI will:

  • Simulate a chain-of-thought discussion between domain experts
  • Ask clarifying questions (in batches)
  • Suggest external data it might need via MCPs

Answer these questions in full so it can proceed with high accuracy.

🔍 3. Review Proposed Codebase Changes

The AI will summarize:

  • Files to be changed or created
  • What will be done
  • Why each change matters
  • Any assumptions made

At this stage: no code is written. You’re expected to review and provide feedback.

✍️ 4. Collaborate on Iteration

You can:

  • Ask the AI to rephrase or simplify any change
  • Ask for alternative implementations
  • Add constraints (e.g., limit new files, avoid new deps)

Once you’re aligned on the changes, give the AI the green light to generate the plan file.

📝 5. Receive the Plan File

The AI will generate:
plans/PLAN_<ISSUE_CODE>.md

This file contains:

  • Implementation phases
  • Actionable steps per file
  • Design rationales
  • Edge case notes
  • Checklists and commit summaries

You can now hand this to another AI agent (or dev) to implement.

🛑 What Not To Do

  • Don’t ask the AI to implement code during planning
    It will refuse until the plan is finalized.

  • Don’t skip the Q&A loop — context matters.

  • Don’t change the output file path/name convention unless you’ve updated the prompt.

📎 Additional Notes

  • The AI will mock external APIs only if necessary — and it will clearly label them.
  • The AI follows the project’s commit message style and Git policies.
  • You can version the plan file as needed (e.g., PLAN_ADE-101_v2.md for iterations).

📣 Questions?

If anything is unclear, just type your question into the coding agent while using the prompt. The AI will either answer or escalate for clarification.