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/drmoisan/drm-copilot --skill benchmark-baselines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。
Admit one new item into a running parallel run — preparation via a preparation-mode child orchestrator run, conflict-edge computation against all items including in-flight ones, and the admission decision that either places the item in the current cohort or defers it and recolors the unstarted subgraph. Appends exactly one mutations[] entry. In-flight items are never moved.
Execute a prepared parallel run for the parallel-orchestrator agent — cohort scheduling under a max_concurrency cap, per-item fan-out onto isolated worktrees branched from origin/main, per-item merge to main after durably confirming CI green, worktree cleanup, and the generated parallel-status.md projection. There is no integration branch and no final integration pull request.
Prepare a set of thematically unrelated items for concurrent execution before any execution begins - item intake over issue numbers and potential-entry paths, concurrent preparation-mode child orchestrator delegations, blast-radius computation and V1-V3 validation, cohort seeding with a recomputation-parity check, run-manifest and planner-checkpoint authoring, and the parallel-orchestrator kickoff artifact.
基于 SOC 职业分类
正在显示 SKILL.md
| name | benchmark-baselines |
| description | Benchmark baseline provenance rules for performance regression gates. |
Source: legacy Claude rule benchmark-baselines.
This rule governs performance baselines used by benchmark regression gates. It exists because a baseline captured on a developer workstation was compared against a windows-latest runner, producing deterministic latency regressions that the benchmark gate could not survive (issue #26, PR #30).
Performance baselines must be captured in the same runner environment class against which they are compared. A baseline captured on a developer workstation must not be committed for comparison against a CI runner.
A baseline whose HostEnvironmentInfo.ProcessorName is the literal string "Unknown processor" is rejected. This value indicates the baseline was captured in an environment where the processor could not be identified (typically a virtualized or developer workstation), which violates runner-environment parity.
HostEnvironmentInfo.ProcessorName == "Unknown processor".Every committed baseline file MUST have a sibling baseline.provenance.json in the same directory. The provenance file records, at minimum:
runner_class — the runner environment class that produced the baseline (for example windows-latest).
host_signature — a stable signature of the host (for example a hashed or labeled description of the CPU/core configuration).
workflow_run_url — the URL of the workflow run that produced the baseline.
Tooling MUST reject a baseline that has no sibling baseline.provenance.json.
The rejection is a Blocking finding; the baseline must be recaptured with provenance recorded.
scripts/benchmarks/Test-BaselineProvenance.ps1 enforces both rejection conditions above and accepts a runner-captured baseline whose ProcessorName is a real processor and whose sibling baseline.provenance.json is present.modified-workflow-needs-green-run (see .agents/skills/feature-review-workflow/SKILL.md) provides a second line of defense: a diff under scripts/benchmarks/** is Blocking unless a green workflow run against the branch head is present in remediation inputs.