Developer cuts AI hook latency from 5 seconds to near-zero using 60-second shell cache

A developer writing hooks for an AI coding assistant found that a cost-monitoring script called on every user prompt was causing delays averaging 819ms, with worst-case waits exceeding five seconds. The bottleneck was a Node.js CLI tool invoked on each prompt, which carried a high startup cost. To fix this, the developer built a 60-second file-based cache in a single shell script, skipping the expensive external call whenever a fresh cached result existed. A critical design detail was writing the cache even when no warning was present, ensuring the common no-alert path was just as fast as the alert path. The approach also handles failure cases by writing an empty cache file, preventing repeated retries on every prompt after an error.