Plan Agent

What This Demonstrates

Copilot’s Plan Agent / Planning Mode generates structured implementation plans, migration strategies, and execution-ready TODO lists — grounded in the actual codebase.

Prerequisites

• VS Code with Copilot Chat open
• Repository loaded (Codespaces or local clone)

Prompts

Prompt A — Feature Planning

Create an implementation plan for adding a `wishlist-service` to this monorepo.
Keep it consistent with existing patterns in `services/*` (Express + TypeScript),
include API endpoints, database schema ideas, events, tests, Docker, Helm, and CI
updates. Output phased steps with clear acceptance criteria.

Why this works

Copilot analyzes the existing service structure and generates a plan that follows the established Express + TypeScript + Prisma patterns. It picks up on conventions like the success/data/error/meta API response format and per-service database ownership.

What to look for:

• Phased approach matching existing services (controllers → services → repositories)
• Database schema following Prisma conventions
• Events for inter-service communication (e.g., wishlist.item.added)
• Helm chart and CI workflow updates

---

Prompt B — Migration Planning

Build a phased plan to introduce API gateway rate limiting and request tracing
across all backend services. Include risk areas, rollout strategy, and rollback
steps.

Why this works

Cross-cutting concerns require understanding all services. Copilot maps the service topology and proposes a safe, incremental rollout with rollback strategies.

What to look for:

• Per-service impact analysis
• Rollout phases (dev → staging → production)
• Rollback procedures tied to Helm and Kubernetes

---

Prompt C — Execution-Ready Plan

Turn the wishlist plan into an actionable TODO list with dependencies and the
first 3 tasks I should implement now in this repo.

Why this works

Transitions from high-level planning to concrete, immediately actionable work items — useful for sprint planning or rapid prototyping.

What to look for:

• Dependency graph between tasks
• Specific file paths and commands
• Clear “done” criteria for each task

Follow-Up Prompts

• “Estimate complexity for each task using T-shirt sizes.”
• “Which of these tasks could be parallelized across two developers?”
• “Generate the Prisma schema for the wishlist-service database.”