/unity-profile

name: unity-profile
description: "Deep profiling session — captures frames via MCP, analyzes CPU/GPU timing, memory snapshots, rendering stats, and provides optimization recommendations."
user-invocable: true
args: focus

/unity-profile — Deep Profiling Session

Run a deep profiling session. Focus: **$ARGUMENTS**

Workflow

Use the `unity-optimizer` agent to:

Step 1: Capture

manage_profiler action:"start_session"     → begin recording
manage_profiler action:"get_frame_timing"  → CPU and GPU frame times
manage_profiler action:"get_counters"      → specific perf counters
manage_profiler action:"memory_snapshot"   → detailed memory breakdown
manage_graphics action:"get_rendering_stats" → draw calls, batches, triangles, set passes
manage_physics  action:"get_stats"         → physics step time, contacts, bodies

Step 2: Analyze

Present a profile report:

**Frame Timing:**

• CPU frame time: Xms (target: <16.6ms for 60fps)
• GPU frame time: Xms
• Bottleneck: CPU / GPU / balanced

**Rendering:**

• Draw calls: X (budget: <100 mobile, <2000 desktop)
• Batches: X (SRP Batcher efficiency)
• Triangles: X
• Set pass calls: X

**Memory:**

• Total: X MB
• Textures: X MB
• Meshes: X MB
• Audio: X MB
• Scripts: X MB

**Physics:**

• Physics step time: X ms
• Active rigidbodies: X
• Contacts: X

Step 3: Recommend

Provide specific, actionable optimization recommendations ranked by impact: 1. Highest impact fix 2. Second highest 3. ...

Each recommendation includes: what to change, why, expected improvement.