profile

If $ARGUMENTS is empty or not provided:

Usage: /profile <target>

Examples:

• /profile ./pkg/auth/ — profile all benchmarkable code in a package
• /profile ProcessOrder — profile a specific function
• /profile . — profile the current package

Workflow: baseline → CPU profile → memory profile → trace (if concurrent) → rank bottlenecks → optimize one at a time and verify with benchstat → final before/after comparison.

Ask: "What file, function, or package would you like me to profile?"

---

If $ARGUMENTS is provided:

Profile Go code, identify bottlenecks, apply optimizations, and verify improvements with statistical rigor. Systematic profile → isolate → optimize → verify cycle.

Loop Initialization

!if [ ! -x "${CLAUDE_PLUGIN_ROOT}/scripts/setup-loop.sh" ]; then echo "ERROR: Plugin cache stale. Run /gopher-ai-refresh (or refresh-plugins.sh) and restart Claude Code."; exit 1; else "${CLAUDE_PLUGIN_ROOT}/scripts/setup-loop.sh" "profile" "COMPLETE"; fi

Configuration

• Target: $ARGUMENTS (file, function, or package)

IRON LAW: NEVER OPTIMIZE WITHOUT PROFILING DATA FIRST

Do not guess. Do not "just try changing X." Every optimization must be preceded by profiling data showing WHERE the bottleneck actually is. Measure before AND after every change.

Phase 1: Environment & Baseline

1. Check tooling:

   which benchstat 2>/dev/null && echo "benchstat: available" || echo "benchstat: NOT FOUND"
   go version
   

   Install benchstat if missing: go install golang.org/x/perf/cmd/benchstat@latest

2. Locate target code — read it, identify hot path, note concurrency and I/O patterns.

3. Find or create benchmarks:

   rg -l 'func Bench' --glob '*_test.go' ./target/path/ 2>/dev/null
   

   If none found: generate table-driven benchmarks with b.ReportAllocs(), realistic inputs, sink variables, b.StopTimer()/b.StartTimer() around setup. Follow the /bench patterns.

4. Establish baseline:

   go test -bench=. -benchmem -count=6 -run=^$ ./target/path/ 2>&1 | tee .profile-baseline.bench
   benchstat .profile-baseline.bench
   

   If variance >5%, increase -count or investigate environment noise.