ptc_runner

Run the full release process for PtcRunner - validates, bumps version, commits, tags, and pushes

Guide for designing, running, and interpreting LLM benchmark experiments — prompt ablation, statistical analysis of pass rates, per-turn interaction metrics, and data-leakage prevention. Use this skill when the user is: running benchmark tests against LLM prompts or configurations, comparing prompt variants (A/B testing prompts), analyzing benchmark results for statistical significance, designing test suites for LLM behavior, investigating per-turn LLM interaction quality, or asking whether sample sizes are sufficient. Also use when the user mentions ablation, pass rate, confidence intervals, Fisher exact test, or prompt optimization in a benchmarking context.