Writing Implementation Plans
Core Principle
A good plan makes implementation obvious. If someone has to guess, the plan is incomplete.
When to Use
Before implementing any multi-step feature. Before delegating to subagents via spawn_agent. Even when the task seems simple, because assumptions cause bugs.
Plan Structure
Every plan must include:
- Goal — One sentence. What does "done" look like?
- Current state — What exists now? What files are relevant?
- Steps — Each step is one action (2-5 minutes of focused work).
- Files to change — Exact paths relative to repo root.
- Verification — How to confirm each step worked.
- Risks — What could go wrong? What assumptions are we making?
Bite-Sized Task Granularity
Each step must be a single action:
- "Write the failing test" is a step.
- "Implement the feature" is NOT a step. Break it down.
- "Add the struct and write tests" is NOT a step. That is two steps.
Output Format
Save the plan as markdown. If using Aixlarity's command system:
# .aixlarity/commands/plan/feature-name.toml
description = "Plan for implementing feature X"
prompt = """Study @{docs/architecture.md} and create an implementation plan for {{args}}."""