Performance Optimization Playbook

Instructions

You are helping a developer improve code performance.

1. Clarify the symptom

   • What is slow (endpoint, job, UI screen, query)?
   • What changed recently (deploy, config, dependency update, data volume)?
   • What is the target (p95 latency, throughput, memory)?

2. Choose the workflow

   • Use the chat workflow when the problem is contained to a small snippet (roughly tens of lines) and you want fast complexity analysis and a rewrite proposal.
   • Use Claude Code when the issue spans multiple files, requires coordinated refactors, or you want changes applied and tests run.

3. Function-level analysis (chat workflow)

   • Identify algorithmic complexity (e.g., nested loops, repeated scans).
   • Look for I/O inside loops (database/network calls).
   • Propose an improved approach (hash map lookup, batching, caching, reduced allocations).
   • Provide an optimized version that preserves behavior.

4. Project-wide optimization (Claude Code workflow)

   • Start in the most performance-critical directories (e.g., api/, core/).
   • Ask for a hypothesis about the regression and where to measure.
   • Apply the smallest safe change first.
   • Re-run benchmarks/tests after each change.

5. Document the result

   • What bottleneck was found?
   • What change fixed it?
   • What measurement proves improvement?

Bundled references:

• templates/performance-triage-questions.md
• examples/dashboard-function-optimization.md

Examples

Example 1: Ask for function-level optimization

Use the prompt in examples/dashboard-function-optimization.md.

Example 2: Ask Claude Code to optimize a performance-critical area

In Claude Code, run a request like:

• Optimize this payment processing function and benchmark results

(If you need a fuller prompt structure, use templates/performance-triage-questions.md.)