Skip to content

exercise-11-implementation-prompt.md

Latest commit

 

History

History
101 lines (65 loc) · 3.63 KB

exercise-11-implementation-prompt.md

File metadata and controls

101 lines (65 loc) · 3.63 KB

Exercise 11 — Create an Implementation Prompt File

Duration: 3 minutes
Copilot Feature: Prompt Files (.prompt.md)
Goal: Package your implementation plan into a reusable prompt file that any team member can invoke.

Background

A Prompt File (.prompt.md) in GitHub Copilot is a saved, reusable prompt. Unlike typing a prompt in chat, prompt files:

  • Are stored in .github/prompts/ and accessible to the whole team
  • Can include instructions, context references, and parameterized inputs
  • Can be invoked quickly from the Copilot Chat command palette

The Implementation Plan prompt file you create here captures the standard way to generate a development plan from the FRD for this project — and any future project following the same process.

Step 1 — Understand the Prompt File Format

A .prompt.md file looks like this:

---
name: Prompt Name
description: "When to use this prompt"
---

# Prompt content here

Instructions for Copilot...

The description field is what you see in the command palette when typing /.

Step 2 — Inspect the Pre-Built Prompt File

The workshop already includes a prompt file at .github/prompts/implementation-plan.prompt.md.

Open it and read it. Notice:

  • The YAML frontmatter with name and description
  • Instructions for structuring phases and task tables
  • Rules about referencing FRD IDs and flagging Background Agent candidates

Step 3 — Create a Custom Project-Specific Version

Now create your own version tailored to your chosen tech stack. In Copilot Chat (local agent), send:

Use /create-prompt with name itms-implementation-plan and paste the following content:

This prompt is specifically for the Intelligent Task Management System (ITMS) project.
When invoked, it should:
1. Read #frd.md and #tsd.md
2. Generate a phased implementation plan with our specific tech stack: [YOUR STACK e.g. TypeScript/Express/PostgreSQL]
3. Reference the correct folder structure: src/routes/, src/services/, src/repositories/, src/models/
4. Include database migration tasks using our tooling convention
5. Include tasks for OpenAPI spec generation
6. Flag tasks suitable for background agent execution

Format: phases as H2 headers, tasks as a table with columns: ID | Task | Effort | FRD Ref | Parallel? | Background Agent?

Replace [YOUR STACK] with your choice from Exercise 09.

Step 4 — Use the Prompt File

Test your new prompt file:

  1. In Copilot Chat, type / — the command palette opens
  2. Type itms — you should see itms-implementation-plan appear
  3. Select it and press Enter
  4. Copilot executes the prompt and generates a fresh implementation plan

Step 5 — Compare the Two Outputs

Notice the difference between:

  • The generic .github/prompts/implementation-plan.prompt.md (for any project)
  • Your project-specific itms-implementation-plan.prompt.md (ITMS-specific folder conventions, stack, tools)

This is why teams maintain both generic prompt libraries and project-specific ones.

Key Takeaway

Prompt files turn one-time instructions into team assets. Instead of one person knowing "the right way to ask Copilot for an implementation plan," the whole team has a /itms-implementation-plan command they can invoke consistently. This is how you scale Copilot adoption across a team.

Next: Exercise 12 — Build APIs with Local Agent

🟡 Optional stop available: If your GitHub repository and Personal Access Token are configured, try Exercise 17 — Create GitHub Issues via MCP before continuing to Exercise 12.