Agent Skills: Simplify Code — Reduce Unnecessary Complexity

Reduce complexity of over-engineered code. Identify unnecessary abstractions, remove dead code, flatten deep nesting, and simplify logic while preserving behavior.

UncategorizedID: vudovn/antigravity-kit/simplify-code

Install this agent skill to your local

pnpm dlx add-skill https://github.com/vudovn/ag-kit/tree/HEAD/.agents/skills/simplify-code

Skill Files

Browse the full folder contents for simplify-code.

Download Skill

Loading file tree…

.agents/skills/simplify-code/SKILL.md

Skill Metadata

Name
simplify-code
Description
Reduce complexity of over-engineered code. Identify unnecessary abstractions, remove dead code, flatten deep nesting, and simplify logic while preserving behavior.

Simplify Code — Reduce Unnecessary Complexity

The best code is the code you don't have to write. The second best is the code anyone can read.

Core Principle

Complexity is a cost. Every abstraction, every indirection, every clever pattern
adds cognitive load. Simplify ruthlessly unless complexity serves a clear purpose.

Simplification Checklist

1. Unnecessary Abstractions

| Smell | Simplification | |---|---| | Wrapper class that just delegates | Remove wrapper, use the inner class directly | | Factory that creates only one type | Replace with direct constructor | | Strategy pattern with one strategy | Replace with simple function | | Interface with one implementation | Remove interface, use the class | | Abstract class with one child | Merge into the child class | | Config object for 2 values | Use function parameters |

2. Dead Code

| Smell | Action | |---|---| | Unused imports | Remove | | Unreachable branches | Remove (check tests first) | | Commented-out code | Remove (it's in git history) | | Unused variables/functions | Remove | | TODO comments older than 6 months | Remove or create issue | | Feature flags for launched features | Remove flag, keep the code |

3. Deep Nesting

// ❌ Before: 4 levels deep
function process(data) {
  if (data) {
    if (data.items) {
      for (const item of data.items) {
        if (item.active) {
          doSomething(item)
        }
      }
    }
  }
}

// ✅ After: Early returns + filter
function process(data) {
  if (!data?.items) return

  data.items
    .filter(item => item.active)
    .forEach(doSomething)
}

4. Over-Parameterized Functions

// ❌ Before: 8 parameters
function createUser(name, email, age, role, dept, active, verified, avatar) { }

// ✅ After: Object parameter
function createUser(opts: CreateUserOpts) { }

5. Premature Optimization

| Smell | Simplification | |---|---| | Custom cache for <100 items | Remove cache, measure first | | Memoization on cheap functions | Remove memo | | Lazy loading for small modules | Use direct import | | Complex state machine for 3 states | Use simple if/else or switch |


Simplification Protocol

Step 1: Identify Complexity

- Count nesting levels (target: ≤3)
- Count function parameters (target: ≤4)
- Count lines per function (target: ≤30)
- Count abstractions per feature (target: ≤2)
- Check for dead code (target: 0)

Step 2: Verify Understanding

Before simplifying, ensure you understand:

  • What the code does (not what it looks like it does)
  • Why it was written this way (maybe there's a reason)
  • What tests cover it (simplification must not break tests)

Step 3: Simplify Incrementally

1. Remove dead code first (safest)
2. Flatten nesting with early returns
3. Inline trivial abstractions
4. Merge related functions
5. Simplify data structures

Step 4: Verify Behavior Preserved

npm run test    # All existing tests still pass
npm run build   # Still compiles

When NOT to Simplify

| Situation | Why Keep Complexity | |---|---| | Performance-critical hot path | Optimization may look complex but is necessary | | Required by framework/library | External constraints | | Explicitly requested pattern | User chose this architecture | | Will need extension soon | Abstraction prepares for known growth |

Ask first: "This pattern seems over-engineered. Should I simplify it, or is there a reason for the abstraction?"