skill-cold-start-warm-path-optimizer

Analyze cold-start latency by computing per-skill p50 and p95 metrics from execution logs, comparing cold versus warm paths to identify prewarm candidates and generate a prioritized prewarming plan.

Retry storms occur when cold starts inflate latency and trigger repeated invocations. Prevent them by generating never_auto_invoke lists for risky skills and prewarming high-latency candidates identified through log analysis.

Generate a prewarm policy by reading execution logs to compute cold and warm path p50/p95 metrics, then scoring skills to produce prewarm_candidates and never_auto_invoke lists exported as machine-readable analysis JSON.

Yes, you can disable auto-invoke by analyzing execution logs to compute cold-start penalties, then generating a never_auto_invoke list for rare or risky skills to safeguard against latency spikes and retry storms.

Compare first-run versus warm-run performance by collecting execution logs and computing per-skill p50/p95 latency for cold and warm paths, revealing latency gaps that guide prewarm candidate selection and auto-invoke policy.

Yes, execution logs are required to compute per-skill p50/p95 cold and warm path metrics, identify cold-start penalties, and generate a prioritized prewarm plan with machine-readable analysis JSON output.