Purpose
Use this skill to evaluate and improve modularity: keep responsibilities tightly grouped (cohesion), keep dependencies small and stable (coupling), and keep boundaries explicit.
This skill fixes one common failure mode: judging “module quality” only at a high level. Here, you always evaluate at the lowest meaningful unit (function / class / small module) and use the worst rating as the truth.
When to use
Use this skill when:
- You are unsure whether a change made a module “too mixed” or “too entangled”.
- You want to refactor to reduce change ripple, without rewriting everything.
- You are reviewing an AI-written change and need an objective explanation and a concrete fix plan.
How to use
-
Open
references/modularity.md. Select 1–3 relevant headings and cite them by heading name in your reasoning. -
Decide the evaluation unit (usually a single function, a class, or a small file section). List the units you will judge.
-
For each unit:
- Rate cohesion at the worst level present.
- Rate coupling at the worst level present.
- Identify boundary leaks (external types, vendor exceptions, “framework objects” flowing inward).
-
Propose the smallest fix that improves the worst rating: split time/procedural cohesion into smaller functional units, isolate “translation glue” at boundaries, introduce a thin wrapper, or restructure data flow.
-
Re-check: does the change reduce reader confusion and reduce future change ripple?
Output expectation
- Provide a short rating summary (worst level + reason) and a fix plan with minimal diffs.
- If you propose a larger change, justify it with: reduced ripple, simpler tests, and clearer boundaries.
