.agents/skills/baseline-regression-governance/SKILL.md
管理数值基线回归的全流程规范(基线生成、存储、测试校验、CI门禁与变更准入),适用于固定点/采样点回归、防止版本更新引入数值漂移。关键词:baseline, regression, fixed points, rtol, atol, smoke, nightly, PNJL, transport
npx skillsauth add w5851/Julia_RelaxTime baseline-regression-governanceInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
本技能用于在仓库内建立并维护“数值基线回归治理”机制,确保:
当出现以下任一场景时,应启用本技能:
tests/baselines/<domain>/tests/baselines/relaxtime/baseline_transport_fixedpoints_v1.csvtests/unit/<domain>/tests/unit/relaxtime/test_transport_legacy_models_bridge_smoke.jlscripts/dev/scripts/dev/export_transport_fixedpoint_baseline.jlbaseline_<feature>_<scope>_vN.csv
baseline_transport_fixedpoints_v1.csvexport_<feature>_baseline.jl
export_transport_fixedpoint_baseline.jltest_<feature>_baseline_smoke.jl(或并入现有 smoke)基线 CSV 至少应包含:
T, mu, xi);eta, sigma, zeta);建议:
%.6f / %.16e);rtol=8e-2, atol=1e-6);isapprox(actual, expected; rtol=..., atol=...);eta/sigma/zeta);legacy vs models)与“基线校验”两条线。unit-smoke,避免过多碎片 workflow。任何基线文件改动都必须同时提交:
禁止:
scripts/dev/;tests/baselines/<domain>/;tests/unit/<domain>/ 写入基线校验;tests/baselines/relaxtime/baseline_transport_fixedpoints_v1.csvscripts/dev/export_transport_fixedpoint_baseline.jltests/unit/relaxtime/test_transport_legacy_models_bridge_smoke.jlUNIT_FILES='relaxtime/test_transport_legacy_models_bridge_smoke.jl'julia --project=. tests/unit/runtests.jl新增其他功能基线时,按同一模式扩展:
tests/baselines/pnjl/、tests/baselines/<new-domain>/baseline_pnjl_<feature>_v1.csvscripts/dev/export_pnjl_<feature>_baseline.jltests/unit/pnjl/test_<feature>_baseline_smoke.jl核心原则:
development
Academic and technical writing style guidance focused on clarity, rigor, and human-like narrative flow. Use when drafting papers, polishing manuscripts, rewriting AI-like prose, or improving argument coherence and citation-aware structure.
tools
Prefer semantic symbol tools for precise navigation and refactoring. Use vscode_listCodeUsages and vscode_renameSymbol instead of grep-style text search when changing code symbols.
testing
Structured technical/theoretical research workflow for method selection, model comparison, and implementation-oriented conclusions with validation plans. Use when decisions must map literature evidence to concrete engineering or computational next steps.
testing
OpenAlex-focused scholarly retrieval skill for work/author/institution queries, citation trend snapshots, and topic-level evidence discovery. Use when you need structured literature discovery and bibliometric signals from OpenAlex.