Audit Perf

!cat ~/.claude/skills/audit-workflow.md

Run as the perf dimension. Goal: surface work the runtime does more than once when once suffices, or at O(N) cost when O(1) is available — then fix or gate it.

Find code doing unnecessary work. Patterns that pay:

• N+1 — one call per item where one batched call suffices
• Overfetching — full objects loaded when IDs or counts suffice; unbounded queries w/o pagination
• Wasted recomputation — derived data recalculated every call; in React, re-renders from unstable refs or missing memoization where deps are stable
• Blocking hot paths — synchronous I/O inside request handlers or render paths
• Unbounded growth — in-memory collections w/o a cap

Fix concrete patterns visible in the code, not "might be slow under load." Apply UI lenses only when UI exists; weigh cache layers only when one would land on a hot path.

Auto-fix behavior-preserving moves: batch the loop, add memo w/ stable deps, paginate the query, move blocking I/O off hot path, cap unbounded collection. State per-fix impact concretely — "O(N) queries → O(1)", "re-renders on every keystroke → memoized." Architectural changes (cache tier, index, transport swap, schema denormalization) → sign-off: name the hot path, sketch the change, surface it.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: JHostalek
• Source: JHostalek/dotclaude
• License: CC0-1.0
• Homepage: https://jhostalek.github.io/dotclaude/

Install and usage instructions live in the source repository linked above.