vinvcn/mattpocock-skills-zh-cn
23 skills61.1k installs64.1k starsGitHub
Install
npx skills add https://github.com/vinvcn/mattpocock-skills-zh-cnSkills in this repo
1Grill Me围绕这个计划的每个方面持续追问我 直到我们达成共同理解 沿着 design tree 的每个分支走下去 一次解决一个决策之间的依赖 每个问题都提供你的推荐答案 The grill me skill documents workflows prerequisites and usage patterns grounded in its repository SKILL md Agents should follow the documented steps respect safety and permission notes and cite only capabilities described in the source It triggers on phrases matching the skill description and integrates with the agent toolchain for the tasks outlined in the documentation3.9kinstalls2Grill With Docs围绕这个计划的每个方面持续追问我 直到我们达成共同理解 沿着 design tree 的每个分支往下走 逐一解决决策之间的依赖 对每个问题 都提供你推荐的答案 如果某个问题可以通过探索 codebase 来回答 就去探索 codebase 而不是问我 CONTEXT md docs adr 0001 event sourced orders md 0002 postgres for write model md src 如果根目录存在 CONTEXT MAP md 说明 repo 有多个 contexts 这个 map 指向每个 context 的位置 CONTEXT MAP md docs adr system wide decisions src ordering CONTEXT md docs adr context specific decisions billing CONTEXT md docs adr The grill with docs agent skill provides documented workflows prerequisites triggers and safety guidance from its SKILL md source Agents load it when user requests match the description and follow step by step instructions without inventing capabilities It integrates with standard agent tooling for the tasks inputs outputs and failure modes described in the repository documentation3.7kinstalls3Setup Matt Pocock SkillsIssue tracker issues 存放在哪里 默认 GitHub 也原生支持 local markdown Triage labels 五个 canonical triage roles 使用的字符串 Domain docs CONTEXT md 和 ADRs 的位置 以及读取它们的消费规则 这是 prompt driven skill 不是确定性脚本 先探索 展示发现 和用户确认 然后再写入 git remote v 和 git config 这是 GitHub repo 吗 是哪一个 repo 根目录的 AGENTS md 和 CLAUDE md 是否存在 其中是否已经有 Agent skills section repo 根目录的 CONTEXT md 和 CONTEXT MAP md docs adr 以及任何 src docs adr 目录 docs agents 这个 skill 之前是否已经输出过内容 scratch 表明已经在使用 local markdown issue tracker 约定 总结现有什么 缺什么 然后带用户逐一完成三个决策 每次只展示一个 section 拿到用户回答后再进入下一个 不要一次性倾倒全部三个 假设用户不知道这些术语是什么意思 每个 section 先用简短说明开头 它是什么 为什么这些 skills 需要它 选择不同选项会改变什么 然后展示选项和默认值 The setup matt pocock skills agent skill provides documented workflows prerequisites triggers and safety guidance from its SKILL md source Agents load it when user requests match the description and follow step by step instructions without inventing capabilities It integrates with standard agent tooling for the tasks inputs outputs and failure modes described in the repository documentation3.5kinstalls4Tdd核心原则 测试应该通过 public interfaces 验证行为 而不是验证 implementation details 代码可以完全改变 测试不该因此改变 好测试 偏 integration style 它们通过 public APIs 执行真实 code paths 它们描述系统做_什么_ 而不是_怎么做_ 好测试读起来像 spec 比如 user can checkout with valid cart 清楚说明已有能力 这些测试能经受 refactor 因为它们不关心内部结构 坏测试 和 implementation 绑定 它们 mock 内部 collaborators 测试 private methods 或通过外部手段验证 例如直接查询 database 而不是使用 interface 警示信号是 你 refactor 了 行为没变 但测试坏了 如果重命名内部函数导致测试失败 那些测试测的是 implementation 不是 behavior 示例见 tests md tests md mocking 指南见 mocking md mocking md 不要先写所有测试 再写所有实现 这是 horizontal slicing 把 RED 当成 写所有测试 把 GREEN 当成 写所有代码 The tdd agent skill provides documented workflows prerequisites triggers and safety guidance from its SKILL md source Agents load it when user requests match the description and follow step by step instructions without inventing capabilities It integrates with standard agent tooling for the tasks inputs outputs and failure modes described in the repository documentation3.5kinstalls5Handoff写一份 handoff document 总结当前对话 让 fresh agent 可以继续工作 保存到用户 OS 的 temporary directory 而不是当前 workspace 在文档中包含一个 suggested skills section 建议 agent 应 invoke 的 skills 不要重复已经捕获在其他 artifacts PRDs plans ADRs issues commits diffs 中的内容 改用 path 或 URL 引用它们 Redact 任何 sensitive information 例如 API keys passwords 或 personally identifiable information 如果用户传入了 arguments 把它们视为下一次 session 重点的描述 并据此调整文档 The handoff agent skill provides documented workflows prerequisites triggers and safety guidance from its SKILL md source Agents load it when user requests match the description and follow step by step instructions without inventing capabilities It integrates with standard agent tooling for the tasks inputs outputs and failure modes described in the repository documentation3.5kinstalls6Improve Codebase Architecture暴露 architecture friction 并提出 deepening opportunities 也就是把 shallow modules 变成 deep modules 的 refactors 目标是 testability 和 AI navigability 在每个建议中精确使用这些术语 语言一致性就是重点 不要漂移到 component service API 或 boundary 完整定义见 LANGUAGE md LANGUAGE md Module 任何有 interface 和 implementation 的东西 function class package slice Interface caller 为正确使用 module 必须知道的一切 types invariants error modes ordering config 不只是 type signature Implementation 内部代码 Depth interface 上的 leverage 小 interface 后面有大量 behaviour Deep 高 leverage Shallow interface 几乎和 implementation 一样复杂 Seam interface 所在的位置 可以不原地编辑就改变 behaviour 的地方 用这个词 不用 boundary Adapter 在 seam 处满足 interface 的具体东西 Leverage callers 从 depth 获得的东西 Locality maintainers 从 depth 获得的东西 change bugs knowledge 集中在一个地方 Deletion test 想象删除这个 module 如果复杂性消失 它只是 pass through 如果复杂性在 N 个 callers 中重新出现 它就在发挥价值 The interface is the test surface One adapter hypothetical seam Two adapters real seam 这个 skill 会参考项目的 domain model Domain language 为好的 seams 命名 ADRs 记录 skill 不应重新争论的决策3.4kinstalls7PrototypePrototype 是 用来回答一个问题的 throwaway code 问题决定形状 先识别正在回答哪个问题 来自用户 prompt 周围代码 或在用户在线时直接询问 Does this logic state model feel right LOGIC md LOGIC md 构建一个很小的交互式 terminal app 推动 state machine 跑过纸面上难以推理的 cases What should this look like UI md UI md 在单一路由上生成几种差异很大的 UI variations 并通过 URL search param 和浮动底栏切换 这两个分支会产出非常不同的 artifacts 选错会浪费整个 prototype 如果问题确实模糊且用户不可达 默认选择更匹配周围代码的分支 backend module logic page 或 component UI 并在 prototype 顶部说明假设 1 从第一天就是 throwaway 并明确标记 Prototype code 要靠近它实际会被使用的位置 放在被 prototype 的 module 或 page 旁边 这样上下文清楚 但命名要让随手读代码的人看出它是 prototype 不是 production 对 throwaway UI routes 遵守项目现有 routing convention 不要发明新的顶层结构 2 一个命令即可运行 使用项目现有 task runner 支持的东西 pnpm name python path bun path 等 用户必须能不动脑地启动它 3 默认不持久化 State 保存在内存中 Persistence 是 prototype 要_检查_的东西 不该成为依赖 如果问题明确涉及 database 就用 scratch DB 或带有清晰 PROTOTYPE wipe me 名称的本地文件 4 跳过 polish 不写 tests 不做超过 能跑起来 所需的 error handling 不做 abstractions 重点是快速学到东西 然后删掉 5 暴露 state 每次 action logic 或每次 variant switch UI 后 打印或渲染完整相关 state 让用户看到发生了什么变化 6 完成后删除或吸收 当 prototype 回答了问题 要么删掉它 要么把验证过的决策折进真实代码 不要让它在 repo 里腐烂3.4kinstalls8TriageThe triage skill moves issues through a state machine with category roles bug and enhancement plus state roles needs-triage, needs-info, ready-for-agent, ready-for-human, and wontfix. Maintainer invocations via slash triage show unlabeled, needs-triage, and needs-info buckets oldest first, then triage specific issues with context gathering, category and state recommendations, bug reproduction attempts, and optional grill-with-docs sessions. Outcomes publish agent brief comments for ready-for-agent, human-only briefs for ready-for-human, triage notes for needs-info, and wontfix with out-of-scope docs for enhancements. Every triage comment must start with an AI-generated disclaimer. State transitions return needs-info issues to needs-triage after reporter replies. References include AGENT-BRIEF.md and OUT-OF-SCOPE.md for persistent brief structure and rejection knowledge base. Use when creating issues, triaging bugs, reviewing feature requests, or preparing issues for AFK agents. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions.3.3kinstalls9Git Guardrails Claude CodeThe git-guardrails-claude-code skill sets up a PreToolUse hook that intercepts and blocks dangerous git commands before Claude executes them. Blocked patterns include all git push variants including force, git reset --hard, git clean -f and -fd, git branch -D, and git checkout dot or git restore dot. Setup asks project versus global scope, copies block-dangerous-git.sh to .claude/hooks with chmod plus x, and merges hook config into .claude/settings.json or global settings without overwriting existing hooks. Blocked commands return exit code 2 with stderr message that Claude lacks permission. Customization step allows adding or removing blocked patterns in the copied script. Verification pipes sample JSON tool_input for git push through the script expecting BLOCKED output. Use when preventing destructive git operations, adding git safety hooks, or blocking git push and reset in Claude Code environments.3.3kinstalls10Setup Pre CommitThe setup-pre-commit skill. 在当前 repo 设置 Husky pre-commit hooks，集成 lint-staged (Prettier)、type checking 和 tests。Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing. Detect package manager 检查 (npm)、 (pnpm)、 (yarn)、 (bun)。使用存在的那个。不清楚时默认 npm。 ### 2. Install dependencies 作为 devDependencies 安装： ### 3. Initialize Husky 这会创建 dir，并向 package.json 添加 。 ### 4. Create 写入这个文件（Husky v9+ 不需要 shebang）： **Adapt**：把 替换为检测到的 package manager。如果 repo 的 package.json 没有 或 script，就省略对应行并告知用户。 ### 5. Create (if missing) 只有没有 Prettier config 时才创建。使用这些 defaults： ### 7. Verify - [ ] 存在且可执行 - [ ] 存在 - [ ] package.json 中的 script 是 - [ ] prettier config 存在 - [ ] 运行 验证可用 ### 8. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions.3.3kinstalls11Scaffold Exercises创建包含 sections problems solutions 和 explainers 的 exercise directory structures 并确保通过 linting Use when user wants to scaffold exercises create exercise stubs or set up a new course section name scaffold-exercises description 创建包含 sections problems solutions 和 explainers 的 exercise directory structures 并确保通过 linting Use when user wants to scaffold exercises create exercise stubs or set up a new course section Scaffold Exercises 创建能通过 pnpm ai-hero-cli internal lint 的 exercise directory structures 然后用 git commit 提交 Directory naming Sections exercises 下的 XX-section-name 例如 01-retrieval-skill-building Exercises section 下的 XX YY-exercise-name 例如 01 03-retrieval-with-bm25 Section number XX exercise number XX YY Names 使用 dash-case 小写 连字符 Exercise variants 每个 exercise 至少需要这些 subfolders 中的一个 problem student workspace 包含 TODOs solution reference implementation explainer conceptual material 不含 TODOs 创建 stub 时 除非 plan 指定其他 variant 否则默认使用 explainer Required files 每个 subfolder problem solution explainer 都需要一个 readme md 要求 非空 必须有真实内容 即使只有一行 title 也可以 没有 broken links 创建 stub 时 生成带 title 和 description 的最小 readme md Exercise Title Description here 如果 subfolder 有 code3.3kinstalls12Migrate To ShoehornThe migrate-to-shoehorn skill. 将 test files 从 `as` type assertions 迁移到 @total-typescript/shoehorn。Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data. **Gather requirements** - 询问用户： - 哪些 test files 中的 assertions 造成问题？ - 是否在处理大型 objects，但只关心部分 properties？ - 是否需要传入故意错误的数据来测试 error paths？ 2. **Install and migrate**： - [ ] Install: - [ ] 查找 test files 中的 assertions: - [ ] 用 替换 - [ ] 用 替换 - [ ] 添加来自 的 imports - [ ] 运行 type check 验证. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions. The workflow follows the source SKILL.md contract with progressive reference loading, clear trigger phrases, and practical steps developers can apply directly in agent sessions.3.2kinstalls13TeachThe teach skill runs a stateful multi-session teaching workspace where the agent tutors users on a topic using MISSION.md, RESOURCES.md, learning-records, reference HTML, and lesson HTML files. Philosophy separates knowledge from high-trust resources, skills from interactive lessons, and wisdom from practitioner interaction. Lessons are short self-contained HTML files with Tufte-style typography, zone of proximal development targeting, primary source recommendations, and follow-up question prompts. Fluency versus storage strength guidance favors retrieval practice, spacing, and interleaving for durable learning. Use when users want to learn a skill or concept across multiple sessions with structured lesson artifacts. Agents should follow the SKILL.md workflow end to end, grounding classification in documented commands, file paths, prerequisites, and troubleshooting notes rather than improvising steps. Run a multi-session teaching workspace with lessons, learning records, and mission-grounded skill tutoring. Invoke when User asks to learn a topic, wants teaching lessons, or mentions teach skill workspace. Best for Agents tutoring users on technical topics needing durable spaced lea.2.8kinstalls14Qaqa is an interactive QA skill for developers who want to report bugs conversationally instead of writing issue templates by hand. During a session, the agent listens to the reporter, asks at most two or three short clarifying questions about expected behavior, reproduction steps, and whether the bug is intermittent, then explores the codebase in the background to capture domain language and context. When the report is clear enough, the skill creates user-facing GitHub issues via gh issue create with accurate titles, steps, and project terminology. Developers reach for qa when they say "QA session", want to batch-file defects after manual testing, or need issues written in the repo’s vocabulary without over-interviewing the reporter.1.9kinstalls15Design An Interfacedesign-an-interface is a Claude Code skill in vinvcn/mattpocock-skills-zh-cn based on John Ousterhout's Design It Twice principle from A Philosophy of Software Design: the first interface idea is rarely optimal. The workflow gathers requirements—problem scope, callers, key operations, constraints, and public versus internal boundaries—then launches parallel sub-agents via the Task tool to produce fundamentally different designs for comparison. Developers reach for design-an-interface when designing a new API, exploring module shapes, comparing interface options, or explicitly asking to design it twice before writing implementation code.1.9kinstalls16Edit Articleedit-article is a Chinese-localized skill from vinvcn/mattpocock-skills-zh-cn that edits long-form drafts by splitting content at headings, confirming section intent with the user, then rewriting each block for clarity and flow. The workflow treats information as a directed acyclic graph so section order respects dependencies between ideas. During rewrites, edit-article caps paragraphs at 240 characters to keep technical posts scannable for web readers. Developers reach for edit-article when a README section, changelog narrative, or blog draft exists but reads disjointed after a fast first pass. The skill confirms section boundaries before rewriting, which prevents large structural surprises mid-session. It complements raw drafting skills by focusing on revision, tightening, and publish-ready structure rather than generating topics from scratch.1.9kinstalls17Obsidian Vaultobsidian-vault is a skill from vinvcn/mattpocock-skills-zh-cn for agent operations inside a configured Obsidian vault at a fixed root path with a deliberately flat layout. Organization relies on Title case note names, [[wikilinks]], and index notes such as Skills Index or RAG Index rather than nested folders. Search workflows use find and grep across *.md files by filename or content, or native Grep and Glob tools pointed at the vault path. Creating notes follows conventions: link dependencies at the bottom and aggregate topics through index notes that are simple wikilink lists. Developers reach for obsidian-vault when they want an AI agent to retrieve research memos, append new technical notes, or maintain cross-linked indexes during RAG, skills, or architecture research. The skill assumes shell access to the vault directory and Obsidian-compatible markdown linking syntax.1.9kinstalls18Request Refactor Planrequest-refactor-plan is a Claude Code skill from vinvcn/mattpocock-skills-zh-cn that guides a structured refactor planning workflow before any code changes. It interviews the developer about the problem and solution ideas, explores the repo to verify assertions, surfaces alternative options, defines exact scope, checks test coverage, and breaks implementation into tiny commits following Martin Fowler's incremental refactoring advice. The final deliverable is filed as a GitHub issue with a detailed refactor RFC. Developers reach for request-refactor-plan when they want a safe, reviewable plan instead of jumping straight into large refactors.1.9kinstalls19Writing Shapewriting-shape is a skill from vinvcn/mattpocock-skills-zh-cn for converting raw markdown input—fragments, unstructured prose, or transcripts—into a standalone article file without mutating the source pile. The session begins by reading the entire input, then proposes two or three candidate openings that imply different theses so the user picks a direction. The agent grows the article paragraph by paragraph, debating format choices such as lists, tables, callouts, and quotes at each step. If no output path is given, writing-shape asks once and re-reads the article file before every write to preserve user edits mid-session. Developers reach for writing-shape when internal notes, interview transcripts, or brainstorm dumps need shaping into a technical blog post or documentation narrative. The raw material file stays read-only while all publishing edits land in the separate article document.1.9kinstalls20Ubiquitous Languageubiquitous-language formalizes domain terminology from the current conversation into a consistent DDD-style glossary saved as UBIQUITOUS_LANGUAGE.md. The skill scans nouns, verbs, and concepts, flags ambiguities where one word maps to multiple concepts or synonyms collide, then proposes canonical term choices with opinionated recommendations. Developers invoke it when defining domain models, hardening terminology, or preparing shared language before coding accelerates. The skill sets disable-model-invocation to true, so it runs only on explicit user request rather than automatic agent triggers.1.9kinstalls21Writing Fragmentswriting-fragments is a skill from vinvcn/mattpocock-skills-zh-cn that runs a structured grilling interview to collect writing fragments before any outline or draft exists. The agent interviews around the user's topic, appends claims, vignettes, sharp sentences, and half-thoughts to a single markdown file, and re-reads the file before each write to preserve user edits. writing-fragments explicitly excludes phases, outlines, and structure—only raw material accumulation. Developers who blog or publish technical articles reach for writing-fragments when they have a topic but need ideation before structuring. The first write adds only a working-title H1 with no metadata, TOC, or date headers.1.9kinstalls22Writing Beatswriting-beats is a skill from vinvcn/mattpocock-skills-zh-cn that shapes articles as a journey of beats in choose-your-own-adventure style. The agent presents 2-3 candidate starting beats from raw material, writes only the selected beat to the article file, re-reads from disk, then offers 2-3 next-beat pivot options until the piece ends naturally. Each beat sets a scene, lands a point, asks a question, or inserts aside—one narrative step at a time. Developers with raw fragments from writing-fragments reach for writing-beats when they want narrative assembly rather than argumentative structure. The skill asks for the article save path once if not provided, then loops write-read-offer until completion.1.9kinstalls23Translate Skilltranslate-skill is a Claude Code skill for ai & agent building. It helps solo builders move faster with AI-assisted coding.11installs