Modularize

SkillFiles & storage

Lets your agent split and refactor oversized code files into smaller, well-organized modules.

Use Modularize in Claude, ChatGPT or Ahel Desktop

Free. Sign in, add Modularize and connect your AI. About a minute.

Also: Claude Code · Cursor · Codex

Then ask your AI: use the Modularize skill

Details

Instructions available. Your AI can read the instructions. Execution depends on the setup they require.

Add Ahel to your AI once: Claude, ChatGPT, Cursor, Claude Code or Codex. Then ask it to use this.

ModularizeStart free
About this skill

Automatically refactor, split, and compress codebase files when they exceed size or complexity boundaries. Trigger when the user requests to split a file, modularize code, or refactor a large module, and when files exceed size (~200 lines) or complexity guidelines.

What this skill tells your AI

The instructions your AI receives, as published by yamcodes/arkenv in skills/modularize/SKILL.md and read by Ahel’s review.

A skill for automatically refactoring, splitting, and compressing codebase files when they exceed size or complexity boundaries.

1. Core heuristics & guardrails

Rather than enforcing a strict, rigid line count, utilize ~200 lines as a warning light/signal to evaluate logical cohesion and maintainability.

Heuristics to evaluate before refactoring:

  • Logical Cohesion: Does the file focus on one primary concern? Or does it contain independent blocks?
  • Function/Block Size: Prioritize keeping functions small (under 20-30 lines) first.
  • Nesting/Complexity: High cyclomatic complexity/nesting vs. simple flat lists.
  • Fragmentation Risk: Avoid "jump-to-definition" fatigue by not over-splitting cohesive units.

Guardrails:

  • The "Abort" Condition: Skip refactoring if the file is predominantly a configuration dictionary, list of constants, or single massive switch statement that logically belongs together.
  • API Contract Preservation: Maintain public APIs using barrel files (e.g., index.ts) to re-export split modules, preventing breaking changes for consumers.
  • Dependency Injection: Favor passing context via arguments rather than creating tangled relative imports (e.g., ../../../).

2. Refactoring patterns

A. test files

For large test suites (e.g., splitting a 600+ line test file):

  • Split by Behavior/Topic: Group tests into separate files named after the feature or option under test (e.g. coercion.test.ts, arrays.test.ts, options.test.ts, etc.).
  • Shared Setup: Extract common setup, beforeEach/afterEach hooks, mocks, or helper functions into a small local test utility or shared helper file, but only when duplication is real and significant.
  • Run Focused Tests: After splitting, run the affected test suite files individually before running the wider project check.

B. source modules

For CLI commands, components, or services:

  • Isolate Concerns: Isolate interactive UI prompts, file system adapters, API clients, and terminal reporters into discrete sub-modules.
  • High-Level Orchestrator: Keep the original file purely as a high-level orchestrator that coordinates the sub-modules.
  • Avoid Speculative Abstractions: Avoid creating generic reusable classes/functions before there are multiple active callers. Keep them as local module splits first.

3. Step-by-step workflow

  1. Inspect & Classify: Read the file, identify dependencies/imports, and evaluate heuristics to decide whether to refactor or abort.
  2. Design Boundaries: Formulate a plan for the split. Define clear boundaries, minimal interfaces, and ensure no public contracts are broken.
  3. Incremental Execution: Extract sub-modules or smaller test files in small, logical batches. Check compilation/type-checking at each step.
  4. Update Imports & Exports: Re-wire caller modules, update barrel/index files, and resolve imports cleanly.
  5. Verify: Run the targeted tests, execute typechecking (e.g., pnpm typecheck or pnpm typecheck:packages), run the package build step (e.g., pnpm build or pnpm build:packages) to guarantee ESM/CJS module resolution/exports remain intact, and execute linting and formatting checks (e.g., pnpm fix).

Signals

GitHub stars
145
Forks
6
Last commit
Oct 2026
Advanced
Item type
skill
Key
modularize-yamcodes
Source
github.com/yamcodes/arkenv