Search

PythonReactSEOPDF
GitHub Skill

unity-performance-check

Use before committing code or after major changes to verify performance. Catches expensive operations before they cause frame drops.

Sharks820☆ 0 Repository stars0 Repository forks

Installation

Follow the repository instructions for this skill and your agent. Review the files and dependencies before adding it to your workflow.

View source ↗Guides

Skill instructions

Unity Performance Check

Overview

Verify code performance BEFORE it causes problems in production. Frame drops ruin games.

When to Use

  • Before committing new code
  • After implementing Update/FixedUpdate logic
  • When adding new systems
  • Before major milestones
  • When users report lag

The Quick Check (30 seconds)

Run these searches on changed files:

// RED FLAGS - Fix these immediately:
GetComponent         // In Update? Cache it
Find(                // In Update? Never do this
FindObjectOfType     // Anywhere? Avoid if possible
new List             // In Update? Preallocate
new Dictionary       // In Update? Preallocate
.ToString()          // In Update? Cache or avoid
string +             // In Update? StringBuilder
foreach              // On non-List? May allocate

The Full Check (5 minutes)

Use the unity-performance-profiler agent:

Task: Launch unity-performance-profiler agent with:
"Analyze [file/folder] for performance issues"

Performance Rules

Update Loop (60fps = 16.6ms budget)
Operation Cost Rule
GetComponent ~1ms Cache in Start
Find/FindObjectOfType ~5-50ms Never use
Instantiate ~2-10ms Use object pool
new List/Dictionary ~0.1ms + GC Preallocate
String concat ~0.5ms + GC StringBuilder
LINQ query ~1-5ms + GC Use loops
Physics.Raycast ~0.1ms Batch or limit
Memory (GC spikes cause stutters)
Cause Solution
new in Update Preallocate collections
String operations Cache strings, use StringBuilder
LINQ methods Convert to explicit loops
Boxing (int to object) Use generics
Closures in lambdas Cache delegates
foreach on IEnumerable Use for loops on arrays/lists
VeilBreakers-Specific
System Performance Concern Solution
Combat Damage calc per frame Batch damage events
Brand lookup Switch statement Dictionary lookup
Synergy calc Recalc every frame Cache on party change
UI updates Rebuild every frame Dirty flag pattern

Checklist

Before committing, verify:

  • No GetComponent in Update/FixedUpdate
  • No Find/FindObjectOfType anywhere
  • No allocations in hot paths
  • Collections preallocated
  • Strings not concatenated in loops
  • Events properly unsubscribed
  • Object pooling for spawned objects

If Performance is Bad

  1. Profile first - Don't guess, measure
  2. Find the hotspot - Usually 1-2 methods
  3. Fix the algorithm - O(n²) → O(n) > micro-optimizations
  4. Cache results - Don't recalculate what hasn't changed
  5. Batch operations - Process multiple items together
  6. Use jobs/burst - For heavy math (advanced)

Examples

Read the usage examples in the source repository.

View source ↗

Source history

Indexed on
2026-01-20
Index updated
2026-01-26
Branch
main

Index and sync dates describe this directory. View the source repository for code history.

GitHub ↗

chat Comments (0)

chat_bubble_outline

No comments yet. Be the first to share your thoughts!