Skill: Build Small, Stay Aligned

Purpose

Guide coding agents to deliver the smallest useful solution while avoiding unnecessary code, abstractions, dependencies, and technical debt. Prioritize alignment on the intended outcome before implementation.

Core Rules

1. Align before action. Before changing code, confirm the goal, scope, constraints, and definition of done. Identify assumptions that could change the implementation.
2. Find the atomic solution. Break the request into the smallest independently useful deliverable. Avoid expanding a task into a platform, framework, or general-purpose system unless that is explicitly required.
3. Prefer less machinery. Minimize lines of code, dependencies, abstractions, configuration, and custom infrastructure. Treat simplicity and reduced maintenance as positive outcomes.
4. Use existing tools where possible. Before building something, check whether an existing library, command-line tool, API, or project capability already solves the problem. Prefer a focused configuration or integration over custom machinery.
5. Challenge every component. For each proposed file, service, abstraction, dependency, or test, ask:
   • Is this necessary for the requested outcome?
   • Can an existing tool or simpler approach replace it?
   • What ongoing maintenance does it introduce?
6. Build only what is needed, then stop. Don’t add speculative features, future-proofing, or extra layers beyond the agreed scope. Stop once the definition of done is met.
7. Simplify after building. Review the result for scaffolding, duplicated logic, unnecessary files, and obsolete planning artifacts. Remove what is no longer needed, using version control to recover anything mistakenly removed.
8. Validate proportionately. First confirm the deliverable works using a focused manual check or a suitable subagent. Add appropriate higher-level or integration tests as needed. Don’t generate a large testing framework or extensive test machinery without a clear reason.