What problem does it solve? Profiling and benchmarking are not always available, yet teams still need to know why a project is slow, what scales badly, and where memory goes. This Skill performs a static performance review of a codebase, inferring likely bottlenecks from the shape of the code and the workload it serves, and delivers a ranked, evidence-cited list of findings instead of guesswork. ## Core Features & Use Cases - Workload modeling: Builds a model of what the cost scales with (files, requests, rows) and identifies the main loop and hot paths before reading code. - Deterministic pattern counting: Ships a standard-library-only Python script that counts regex-in-loop, per-call allocations, whole-file reads, locks, parallelism, caches, and limit constants across Rust, Java, JS/TS, Python, Go, and C#. - Ranked findings report: Emits a validated JSON findings file with severity-calibrated bottlenecks, deliberate optimizations, and a synthesized performance posture, rendered into an interactive explorer. - Use Case: Ask why your analysis tool slows down on large repositories; the scanner identifies a quadratic dependency-matching stage, regex compiled per line, and a serial stage between parallel ones, each cited to the exact loop. ## Quick Start Ask the AI to run a performance scan on this repository using the existing _sokrates analysis and report the top likely bottlenecks.
