agent-runtime-cache-benchmark

Diagnose prompt caching reuse by comparing two structured JSON run artifacts to quantify cache hits, latency delta, and hash stability. The tool flags likely cache-break events when system prompts, tool manifests, or history change between runs.

Prompt-cache reuse benchmarking measures whether caching improves performance by comparing cold versus warm agent-run artifacts. You need it when verifying if system prompts, tool manifests, or history affect caching across operator workflow runs.

Yes, you can benchmark prompt caching by feeding two JSON run artifacts into the helper script. It computes cache shares, flags likely cache-break fields, and outputs a local Markdown or JSON report for operator review.

The best way to compare cold and warm agent runs is using a benchmarking script that reads two JSON artifacts, computes cache shares, and flags cache-break fields. It generates a concise Markdown or JSON report locally for review.

Prompt caching fails when system prompts, tool manifests, or history change between runs, breaking hash stability. The benchmarking tool identifies these likely cache-break events by comparing two structured run artifacts and quantifying the cache share delta.

No external dependencies are required to generate a prompt-cache benchmark report locally. The tool uses bundled scripts and references to read two JSON run artifacts, compute cache shares, and output a Markdown or JSON report directly.