菜单
After a refactor that reduced complexity or violations, tighten the ratchet baselines (.xenon.yml, .cognitive_baseline.txt, .coverage_baseline.txt, .file_length_baseline.txt) to lock in the gain
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/bj112233/governor-edr --skill ratchet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。
基于 SOC 职业分类
正在显示 SKILL.md
| name | ratchet |
| description | After a refactor that reduced complexity or violations, tighten the ratchet baselines (.xenon.yml, .cognitive_baseline.txt, .coverage_baseline.txt, .file_length_baseline.txt) to lock in the gain |
| argument-hint | [which gate: xenon|cognitive|coverage|file-length|all] |
| allowed-tools | ["read","grep","glob","edit","exec"] |
| triggers | ["user","model"] |
Tighten the ratchet baselines after a refactor that reduced complexity, increased coverage, or removed dead code. This enforces AGENTS.md §4: "Ratchet protocol: extract one block per commit; never lower threshold below proven CC." The ratchet only tightens — never loosen a threshold below its proven value.
| Gate | Baseline file | Regenerate command |
|---|---|---|
| Cyclomatic Complexity (xenon) | .xenon.yml | manual edit (max-absolute, max-average, max-modules) |
| Cognitive Complexity | .cognitive_baseline.txt | .\.venv\Scripts\python.exe bin\cognitive_complexity_gate.py --regenerate |
| Coverage |
.coverage_baseline.txt |
.\.venv\Scripts\python.exe bin\coverage_gate.py --regenerate |
| File Length | .file_length_baseline.txt | manual edit (remove shrunk/extracted files) |
Determine which gate(s): If $ARGUMENTS specifies a gate, focus on it. Otherwise run all 4 gates to see which improved:
.\.venv\Scripts\python.exe bin\lint-gate.py --fast
Note: --fast skips the coverage gate. To check coverage, run .\.venv\Scripts\python.exe bin\coverage_gate.py separately (it reuses cached coverage.json if fresh <10 min).
For each improved gate, regenerate the baseline:
.\.venv\Scripts\python.exe bin\cognitive_complexity_gate.py --regenerate.cognitive_baseline.txt with the current (smaller) violation set..\.venv\Scripts\python.exe bin\coverage_gate.py --regenerate.coverage_baseline.txt from the cached coverage.json.coverage.json is stale (>10 min since last code change), run --fresh first to force a pytest re-run. WARNING: --fresh runs the full 200+ test suite — takes minutes. Only use if explicitly needed.Total: line is ≥ the old value (ratchet only rises)..file_length_baseline.txt. For each file that was refactored (split/extracted), check its current LLOC:
.\.venv\Scripts\python.exe -c "import sys; sys.path.insert(0,'bin'); from file_length_gate import _count_lloc; print(_count_lloc('path/to/file.py'))".\.venv\Scripts\python.exe -m xenon services/ --max-absolute D --max-average A --max-modules C and check the actual average grade..xenon.yml to tighten max-average to the proven grade..xenon.yml to reflect the new state.Verify the gate still passes with the new baseline:
.\.venv\Scripts\python.exe bin\lint-gate.py --fast
If it fails, the baseline was tightened too aggressively — revert and use a less aggressive value.
Report: Summarize what was tightened:
| Gate | Before | After | Change |
|---|
--regenerate (cache reuse, seconds) over --fresh (full pytest, minutes) unless the cache is stale..\.venv\Scripts\python.exe for all commands — the venv hook blocks bare python.