A quality-over-features release that makes every finding more defensible and every skip more deliberate.
🌍 English
Highlights
- 📚 12-book source coverage matrix — The new
skills/_shared/source-coverage.mdmaps every one of the twelve source books (now including The Art of Unit Testing and How Google Tests Software as first-class citizens) to the concepts it contributes, the nuances it insists on, the false positives it warns against, and the review questions it inspires. Every skill now reads this file before writing findings. - 🛡️ "What Not to Flag" guardrails — R1–R6 and T1–T5 now ship with explicit negative guidance. Deep modules are not shallow modules. Composition roots are not DIP violations. Transaction scripts are not anemic domains. CRUD is not distortion. The goal: cut template-driven over-reporting without dulling the real signal.
- 🧪 Benchmark suite expanded to 43 scenarios — Six new cases explicitly test the model's restraint: deep Go modules, Hyrum's Law warnings, closed-protocol switches, TypeScript composition roots, Python transaction scripts, and risk-shaped legacy Java test portfolios.
- 📐 Docs & validation tightened — README, AGENTS.md, CLAUDE.md, and CONTRIBUTING.md now document the coverage matrix.
node scripts/validate-repo.mjsis promoted to a first-class consistency gate; the validator itself now checks for the twelve-book inventory, the source-coverage file, and the "What Not to Flag" sections.
No breaking changes. Existing .brooks-lint.yaml configs, skill invocations, and short-form commands are untouched.
Install / Upgrade
```bash
/plugin marketplace add hyhmrright/brooks-lint
/plugin install brooks-lint@brooks-lint-marketplace
```
🇨🇳 简体中文
一个"重质量而非重功能"的版本：让每一条 finding 更站得住脚，也让每一次"不 flag"更经得起推敲。
亮点
- 📚 12 本书的源头覆盖矩阵：新增 `skills/_shared/source-coverage.md`，为每本源头书（现在 The Art of Unit Testing 和 How Google Tests Software 已正式进入一级清单）明确列出：贡献了哪些概念、强调了哪些细节、警告了哪些假阳性、应该提出哪些审查问题。所有四个 skill 在写 finding 之前都会先读这个文件。
- 🛡️ "What Not to Flag" 防护栏：R1–R6 与 T1–T5 全部新增了显式的反向指导。Deep module 不是 shallow module；composition root 不是 DIP 违规；transaction script 不是贫血领域；CRUD 不是模型扭曲。目标：在不削弱真实信号的前提下，砍掉模板驱动的过度报告。
- 🧪 基准评测集扩充到 43 个场景：6 个新增场景专门检验模型的"克制"：Go 深度模块、Hyrum's Law 预警、封闭协议 switch、TypeScript 组合根、Python 事务脚本，以及按风险形状裁剪的 Java 遗留测试套件。
- 📐 文档与校验收紧：README、AGENTS.md、CLAUDE.md、CONTRIBUTING.md 同步记录覆盖矩阵；`node scripts/validate-repo.mjs` 被提升为一级一致性闸门，校验脚本本身现在也会检查 12 本书清单、source-coverage 文件和 "What Not to Flag" 段落是否齐全。
无破坏性变更：现有 `.brooks-lint.yaml` 配置、skill 调用方式和短命令均不受影响。
安装 / 升级
```bash
/plugin marketplace add hyhmrright/brooks-lint
/plugin install brooks-lint@brooks-lint-marketplace
```
Full changelog / 完整变更日志: CHANGELOG.md