profiling-methodology-designer

What problem does it solve?

This Skill designs a rigorous profiling investigation when something is slow or memory-hungry and nobody knows where the time or allocations are going. It replaces guesswork with evidence by choosing the right profiler class, the right measurement conditions, and the right narrowing loop before any fix work begins.

Core Features & Use Cases

• Attribution-first profiling design: Selects cross-service trace decomposition, on-CPU sampling, off-CPU blocking analysis, or allocation profiling based on the symptom shape.
• Measurement discipline: Defines warm-versus-cold runs, representative load and data volume, duration, sample count, and profiler overhead budgets so results are meaningful.
• Evidence-based narrowing: Builds a hypothesis loop with ruled-out suspects, baseline profiles, differential comparisons, and a final attribution report with handoffs to the owning specialist skill.
• Use case: A service has p99 latency spikes but no clear culprit, and this Skill designs the investigation that identifies whether the bottleneck is queries, waiting, allocations, or a downstream hop.

Quick Start

Ask the skill to design a profiling plan for the slowest request path, including the likely profiler class, measurement conditions, narrowing steps, and handoff targets.