Claude 代码技能库——高级 NestJS/Node.js 后端工程师，及更多
概览
一个个人的Claude代码技能库。每项技能都教会代理一个领域——如何思考、遵循哪些规则、避免哪些错误——在处理这类代码时。 （更多技能计划——React前端指南、expo移动指南等） 或者手动——技能存在于 ~/.claude/skills/ 中，Claude Code 会自动发现它们： 下一次 Claude Code 会话会自动加载技能。没有重启或安装步骤。 在原位编辑文件。更改将在下一次Claude代码会话中生效。 技能是一个包含至少 SKILL.md 文件的文件夹，里面包含YAML前备件（名称、描述）。将任何现有技能视为模板。 - 保持 SKILL.md 可扫描（≤500行）。深入分析需在参考文献中。- 每个参考文件：简而言之;DR →规则（好坏代码）→反模式→复习清单。- 在前言中写描述，使Claude能可靠地触发技能（具体关键词+“使用时X”框架）。最多1024个字符。- 添加评估/评估。
概述（AI 翻译）
一个个人的克劳德代码技能库。每一项技能 教会智能体一个领域——如何思考，遵循哪些规则，避免哪些错误—— 在处理这类代码时。
本收藏中的技能
| 技巧 | 它的作用 |
|---|---|
nestjs-dev-guidelines |
高级NestJS / Node.js后端工程师：执行学科、文件夹结构、命名、模块设计、代码质量、决策树、API设计、标准响应、分页/筛选/排序、验证、错误处理、异常过滤、安全、认证/RBAC、多租户、数据库设计+ORM模式、迁移、级联规则、管道/拦截/守卫、事件、后台作业、配置、日志、可观察性/追踪、测试、性能、N+1消除、缓存、Swagger文档、AI产品模式（LLM网关、SSE流、使用计量和成本）、代码审查检查表+反模式、现代Nest堆栈、健康/准备/关闭、源代码新鲜性检查、webhooks、文件上传、装饰器/示波器/动态模块、DDD分层/六边形架构（端口/适配器/用例）。SKILL.md + 43个细粒参考文件 + 评估。 |
本页由 AI 翻译，技术细节请以原文为准。
README
以下为原文 README 节选，完整内容请查看原文。
vector-skills
A personal library of Claude Code skills. Each skill teaches the agent a domain — how to think, what rules to follow, what mistakes to avoid — when working on that kind of code.
Skills in this collection
| Skill | What it does |
|---|---|
nestjs-dev-guidelines |
Senior NestJS / Node.js backend engineer: execution discipline, folder structure, naming, module design, code quality, decision trees, API design, standard responses, pagination/filters/sorting, validation, error handling, exception filters, security, auth/RBAC, multi-tenancy, DB design + ORM patterns, migrations, cascade rules, pipes/interceptors/guards, events, background jobs, configuration, logging, observability/tracing, testing, performance, N+1 elimination, caching, Swagger docs, AI product patterns (LLM gateway, SSE streaming, usage metering & cost), code review checklist + anti-patterns, modern Nest stack, health/readiness/shutdown, source-of-truth freshness checks, webhooks, file uploads, decorators/scopes/dynamic modules, DDD layered/hexagonal architecture (ports/adapters/use cases). SKILL.md + 43 granular reference files + evals. |
code-implementation |
Architect-grade workflow for changing code in an existing codebase — features, fixes, refactors, integrations, migrations. Read the real code before writing any, trace the root cause instead of patching symptoms, reuse what exists instead of adding near-duplicates, verify end to end. SKILL.md + 5 reference files (investigation, root-cause, design-review, code-quality, verification) + evals. |
bug-fixing |
Disciplined defect workflow — reproduce, read the real code, find the true root cause, design the minimal correct fix at the right layer, verify it, then hunt for clones of the same defect. Covers crashes, wrong output, flaky tests, regressions, races, leaks, perf degradation, works-locally-not-in-prod. SKILL.md + 3 reference files (root-cause techniques, anti-patterns, report templates) + evals. |
code-quality-check |
Read-only, evidence-backed quality review producing severity-ranked findings, a scorecard, and a merge/ship verdict. For PR/diff review, pre-merge gates, and second opinions on agent- or teammate-written code. Hands off to the implementation workflow rather than writing the fix. SKILL.md + 6 reference files + scripts/repo_signals.py + evals. |
推荐工具
换一个关键词，或者移除筛选条件。
安装
npx skillfish add myatminlu/vector-skills