GLD.SerializerBenchmark
GLD.SerializerBenchmark contains 2 collected skills from leo-gan, with repository-level occupation coverage and site-owned skill detail pages.
Skills in this repository
Review serializer-benchmark results for suspicious sizes/ops/fidelity, research each suspect against the library’s official docs and code examples, fix wrong or ineffective harness client call paths (without breaking serializer isolation), re-bench, and verify. Also covers “official package is optimal but still too slow” cases: research ecosystem high-perf peers (e.g. serde_avro_fast vs apache-avro) when a binary/schema codec lags same-lang JSON. Use when the user runs /review-suspicious-results, says "review suspicious results", "audit benchmark numbers", "check outliers", "too fast / too slow serializers", or "verify client calling against package docs".
End-to-end pre-PR gate for the serializer-benchmark monorepo: refuse master/main, run full test suites, full benchmarks only for languages changed on the branch (detect-changed-langs.sh; override with PREPARE_PR_LANGS / PREPARE_PR_BENCH_ALL), fail on error-CSV regressions for those languages, regenerate analysis for them, update dashboard data (sync-data.py), commit, push, and draft a short PR description. Use when the user runs /prepare-pr, says "prepare a PR", "ready for review", "pre-PR gate", or "full PR validation".