后端审查
- 作者仓库星标 0
- 作者更新于 实时读取
- 作者仓库 skills-registry
- 领域
- 写作
- 兼容 Agent
-
- Claude Code
- Cursor
- Cline
- Codex
- Windsurf
- Gemini CLI
- +20
- 信任分
- 88 / 100 · 社区维护
- 作者 / 版本 / 许可
- @tomevault-io · 未声明 license
- Token 消耗评级
- 低消耗
- 接入复杂程度
- 即装即用
- 是否需要外部 API Key
- 不需要
- 兼容的系统
- 未声明（默认跨平台）
- 底层运行要求
- 无特殊要求
- 文件与系统权限
-
- 只读
- 允许写入 / 修改
- 网络行为
- 仅限本地
- 安装命令数
- 26 条
档案由构建时根据 SKILL.md 与安装命令自动衍生，可能与作者实际意图存在差异。
需要注意： 未限定 allowed-tools，默认拥有全部工具权限。
---
name: engineering-foundations
description: Language-agnostic engineering methodology — DDD, Hexagonal Architecture, TDD, architecture plann…
category: 写作
runtime: 无特殊运行时
---
# engineering-foundations 输出预览
## PART A: 任务判断
- 适用问题：文章、文案、发言稿、润色或结构化表达。
- 输入要求：目标材料、限制条件、期望输出和验收方式。
- 证据边界：围绕“Topics / Gotchas / General Principles”读取原文规则，不把推断写成作者承诺。
## PART B: 执行结果
- **01** 任务判断：确认你的需求是否属于文章、文案、发言稿、润色或结构化表达，并标出输入、限制和预期结果。
- **02** 执行计划：优先按“Topics / Gotchas / General Principles”拆成步骤，说明每一步会读取什么、修改什么、产出什么。
- **03** 交付结果：给出可复制的命令、文件改动、检查清单或内容草稿，并说明如何继续迭代。
- **04** 风险边界：结合 读取文件、写入/修改文件、主要在本地完成、通常不需要额外 API Key 给出执行前确认项。
## Running Rules
- 读取文件、写入/修改文件；主要在本地完成；通常不需要额外 API Key。
- 先小样例验证，再放大到真实任务。
- 交付时同时给结果、检查口径和下一步迭代建议。 先确认触发方式
原文没有稳定的斜杠命令要求。安装验证后通常全局生效，直接在对话里点名这个 Skill 并描述任务即可。
给清楚输入和边界
告诉 Agent 目标文件或材料、期望结果、不可改范围、是否允许联网或执行命令。本 Skill 的权限画像是：读取文件、写入/修改文件。
小样例验证后再放大
先用一个小任务确认它会围绕“Topics / Gotchas / General Principles”工作；涉及文件或命令时，先看 diff、日志、预览或测试结果。
复核后再交付
检查最终产物是否包含明确结果、必要证据和下一步动作；如果输出泛泛而谈，就补充输入、边界和验收标准后重跑。
---
name: engineering-foundations
description: Language-agnostic engineering methodology — DDD, Hexagonal Architecture, TDD, architecture plann…
category: 写作
source: tomevault-io/skills-registry
---
# engineering-foundations
## 什么时候使用
- 用于审阅代码、文档或方案并给出可执行反馈 适合处理文章、文案、润色、翻译、总结和结构化表达，核心价值是把输入、判断、执行、验证和交付边界固定下来，避免 Agent 泛泛回答。 把任务拆成可执行、可检查、可继续迭代的步骤；通常不需要额外…
- 面向文章、文案、发言稿、润色或结构化表达，优先处理能明确输入、步骤和验收标准的工作。
## 需要提供什么
- 目标材料、目录范围、期望结果和不可改动内容。
- 是否允许联网、执行命令、读写文件或调用外部服务。
## 执行规则
- 围绕「Topics / Gotchas / General Principles」组织步骤，不把推断写成作者事实。
- 读取文件、写入/修改文件；主要在本地完成；通常不需要额外 API Key。
- 先跑小样例，确认结果可检查后再扩大任务范围。
## 输出要求
- 给出最终产物、关键证据、验证方式和下一步动作。
- 信息不足时标记 unknown，不编造命令、平台或依赖。 证据边界与执行链路
作者原文负责流程事实；仓库文件负责来源和命令；流狐只补充适用场景、限制和质量判断。
skill "engineering-foundations" {
输入层 -> 用户目标 + 目标文件 + 禁止范围 + 验收标准
上下文层 -> Topics / Gotchas / General Principles
规则层 -> SKILL.md 触发条件 / 执行顺序 / 输出格式
运行层 -> 无特殊运行时 | 读取文件、写入/修改文件 | 主要在本地完成
安全层 -> 通常不需要额外 API Key + 小任务验证 + diff / 日志复核
输出层 -> 可复制结果 + 检查清单 + 下一步迭代
}Recommended model: Opus — This skill involves deep reasoning, architecture decisions, or code review.
Opinionated, language-agnostic engineering methodology. Before applying any topic, read its reference file in reference/.
Topics
| Topic | When to Use | Reference |
|---|---|---|
| TDD | Writing unit tests, e2e tests, fixing failing tests | reference/tdd.md |
| Domain Modeling | Implementing rich domain models with business rules, state transitions | reference/domain-model.md |
| Code Review | Reviewing PRs, validating code quality, self-review | reference/code-review.md |
| Requirements | Gathering requirements, clarifying business logic, defining scope | reference/requirements.md |
| Architecture Planning | Planning features, making architectural decisions, designing solutions | reference/architecture.md |
| Deep Modules | Judging interface-vs-implementation ratio, choosing where to draw module boundaries, classifying dependencies | reference/deep-modules.md |
| ADR Writing | Documenting significant architectural decisions | reference/adr.md |
Gotchas
Claude-specific failure modes in engineering methodology:
- Hardcoding dates in tests — Claude's most persistent bad habit. Always use dynamic date generation (
new Date(), relative offsets). Never'2024-01-15'. - Modifying existing tests to make them pass — When a test fails, Claude's instinct is to "fix" the test. Fix the application code instead. Only modify tests if the requirements changed.
- Mocking internal modules in e2e tests — Claude mocks everything for convenience. In e2e tests, only mock external services (payment APIs, auth providers). Internal services, handlers, guards, and repositories must be real.
- One giant PR with no intermediate commits — Claude builds everything then commits once. Commit after each TDD Green phase.
- Designing from scratch when patterns exist — Claude invents new architectures instead of finding the closest existing implementation in the codebase. Always search for similar code first.
- Presenting only one option — Architecture planning requires 2-3 options with trade-offs. Claude tends to present its preferred approach as the only option.
- ADRs with no "Rejected Alternatives" — Claude writes ADRs that only describe the chosen approach. The rejected alternatives section is the most valuable part.
- Skipping the "what's out of scope" section — Requirements must explicitly define what is NOT being built. Claude omits this, leading to scope creep.
General Principles
| Principle | What to Do |
|---|---|
| Clarify Constraints | Scale? Budget? Existing systems? Timeline? MVP or production? |
| Prioritize Simplicity | Simplest solution that works. "The best code is no code." |
| Security by Design | Least privilege, encryption, input validation, parameterized queries |
| Present Options | 2-3 options with trade-offs. Then recommend one. |
Quick Decision Guide
Need to write/fix tests?
-> Read reference/tdd.md
Need to design a domain model?
-> Read reference/domain-model.md
Need to review code?
-> Read reference/code-review.md
Need to gather or clarify requirements?
-> Read reference/requirements.md
Need to plan a feature or make architecture decisions?
-> Read reference/architecture.md
Need to judge a module boundary, or decide if something is "too shallow"?
-> Read reference/deep-modules.md
Need to document a significant decision?
-> Read reference/adr.mdSource: anpham1925/claude-marketplace — distributed by TomeVault.
先判断是否适合
作者设计意图
作者的方法与取舍
边界和复核