Use this skill whenever working with the personal_agent_eval (pae) benchmark framework in this repository. Covers everything needed to operate the library: running benchmarks with `pae run-eval`, creating and modifying test cases, suites, run profiles, and evaluation profiles, understanding fingerprints and incremental reuse, reading and interpreting evaluation outputs, configuring the LLM judge, and all `pae` CLI commands and flags. Trigger on any of these: "run a benchmark", "create a test case", "add a model to a suite", "configure the judge", "why is it reusing results", "how does scoring work", "how do I re-run a case", "what is a fingerprint", "set up a campaign", "add a deterministic check", "what does each dimension measure", "how does the openclaw runner work", or any question about how this library works.
2026-05-14
