菜单
Review a CV-CUDA operator's BENCHMARK coverage — drivers, layout axis, baselines, the basic-tier floor, row counts, and coverage statistics. Use when asked whether an operator's benchmarks/baselines are complete or to find/fill bench gaps.
默认使用会先检查来源的 Prompt；你也可以切换为直接命令，或下载本地副本。
决定是否安装前，请先阅读 SKILL.md，以及 SkillsMP 当前展示的配套文件。
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
直接命令不会经过审查 Prompt；运行前请先检查来源。
npx skills add https://github.com/CVCUDA/CV-CUDA --skill review-op-bench-coverage命令会保持在同一行。复制前请横向滚动并检查完整内容。
想先保存到本地？可下载 SkillsMP 当前能够提供的文件。
Scaffold a new CV-CUDA operator — a complete, wired, building skeleton — and delegate the implementation to a human or another AI. Use when asked to set up / stub out a new operator without implementing it, or to bootstrap one for someone else to finish. Supports --bare (skip the spec when the definition itself is delegated).
Add a new CV-CUDA operator end-to-end per .agents/guidance/MAKE_OP_GUIDELINES.md, with a deterministically-enforced definition-of-done. Use when asked to create/add a new operator, scaffold one, or verify that a new operator is complete (approved spec, wired scaffold, bit-exact regression tests across the declared support matrix, required layout parity, complement negatives, docs + relnote, benched dtypes, tests that run and pass, optimization-ready).
Verify a new CV-CUDA operator against the deterministic final regression checklist (the /make-op done-gate). Use when asked whether a new operator is complete/done, or to gate it before merge - gold reference, bit-exact coverage across the declared support matrix, required layout parity, complement negatives, docs + relnote, benched dtypes, tests that run and pass, and optimization-readiness.
基于 SOC 职业分类
正在显示 SKILL.md
| name | review-op-bench-coverage |
| description | Review a CV-CUDA operator's BENCHMARK coverage — drivers, layout axis, baselines, the basic-tier floor, row counts, and coverage statistics. Use when asked whether an operator's benchmarks/baselines are complete or to find/fill bench gaps. |
Thin entry point. Run python3 tools/review_op.py <Operator> --domain bench (add --run
for the GPU-run-dependent checks) and interpret per the bench section (BEN-*) of
.agents/guidance/REVIEW_OP_GUIDELINES.md: structural rules + the hard basic-tier minimum floor
(BEN-14) + advisory coverage statistics and RECOMMENDATIONs. Resolve MANUAL items at the cited locations. When fixing
(--fix), add the layout axis and required native/reference layout configs, then recompute row counts - a missing-SKU
baseline gap is reported "requires CI" and never fabricated — and re-run to confirm.
Findings-first.