[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"pack-detail-prompt-engineering-toolkit-zh":3,"seo:pack:prompt-engineering-toolkit:zh":68},{"code":4,"message":5,"data":6},200,"操作成功",{"pack":7},{"slug":8,"icon":9,"tone":10,"status":11,"status_label":12,"title":13,"description":14,"items":15,"install_cmd":67},"prompt-engineering-toolkit","🎯","#B91C1C","stable","稳定","Prompt 工程工具箱","Awesome Prompt Engineering \u002F OpenAI Cookbook \u002F Prompt Architect 27 框架 \u002F Prompt Master + Claude Code prompt-engineer subagent。",[16,28,36,43,52,60],{"id":17,"uuid":18,"slug":19,"title":20,"description":21,"author_name":22,"view_count":23,"vote_count":24,"lang_type":25,"type":26,"type_label":27},509,"1b3fa22b-1246-42e5-a6d4-f5211029f6ef","awesome-prompt-engineering-papers-tools-courses-1b3fa22b","Awesome Prompt Engineering — Papers, Tools & Courses","Hand-curated collection of 60+ papers, 50+ tools, benchmarks, and courses for prompt engineering and context engineering. Covers CoT, RAG, agents, security, and multimodal. Apache 2.0.","Prompt Lab",558,0,"en","prompt","Prompt",{"id":29,"uuid":30,"slug":31,"title":32,"description":33,"author_name":34,"view_count":35,"vote_count":24,"lang_type":25,"type":26,"type_label":27},96,"26b9b7dd-dbe6-41e3-a093-17db5409d739","openai-cookbook-official-prompting-guides-26b9b7dd","OpenAI Cookbook — Official Prompting Guides","Official prompting guides from OpenAI: GPT-5.2, Codex, Meta Prompting, and Realtime API guides. The definitive reference for OpenAI model optimization.","OpenAI",374,{"id":37,"uuid":38,"slug":39,"title":40,"description":41,"author_name":22,"view_count":42,"vote_count":24,"lang_type":25,"type":26,"type_label":27},845,"08f51e3b-33aa-11f1-9bc6-00163e2b0d79","prompt-architect-27-frameworks-expert-prompts-08f51e3b","Prompt Architect — 27 Frameworks for Expert Prompts","Transform vague prompts into structured, expert-level prompts using 27 research-backed frameworks across 7 intent categories. Works with Claude Code, ChatGPT, Cursor, and 30+ AI tools.",310,{"id":44,"uuid":45,"slug":46,"title":47,"description":48,"author_name":22,"view_count":49,"vote_count":24,"lang_type":25,"type":50,"type_label":51},848,"0994566a-33aa-11f1-9bc6-00163e2b0d79","prompt-master-zero-waste-ai-prompt-generator-skill-0994566a","Prompt Master — Zero-Waste AI Prompt Generator Skill","Claude Code skill that generates optimized prompts for 30+ AI tools. Auto-detects target tool, applies 5 safe techniques, catches 35 credit-killing patterns. 4.8K+ stars, MIT license.",243,"skill","Skill",{"id":53,"uuid":54,"slug":55,"title":56,"description":57,"author_name":58,"view_count":59,"vote_count":24,"lang_type":25,"type":26,"type_label":27},807,"15f82b68-ac1f-4e52-84b2-cead8b4cb869","ai-prompt-engineering-best-practices-guide-15f82b68","AI Prompt Engineering Best Practices Guide","Comprehensive guide to writing effective prompts for Claude, GPT, and Gemini. Covers system prompts, few-shot learning, chain-of-thought, and structured output techniques.","Skill Factory",321,{"id":61,"uuid":62,"slug":63,"title":64,"description":65,"author_name":58,"view_count":66,"vote_count":24,"lang_type":25,"type":50,"type_label":51},44,"57eff515-8f7b-4a35-9078-98df47ac2d06","claude-code-agent-prompt-engineer-design-test-prompts-57eff515","Claude Code Agent: Prompt Engineer — Design & Test Prompts","Claude Code agent for designing, optimizing, and testing LLM prompts. Improves accuracy, reduces token usage, and benchmarks results.",242,"tokrepo install pack\u002Fprompt-engineering-toolkit",{"pageType":69,"pageKey":8,"locale":70,"title":71,"metaDescription":72,"h1":13,"tldr":73,"bodyMarkdown":74,"faq":75,"schema":91,"internalLinks":101,"citations":114,"wordCount":127,"generatedAt":128},"pack","zh","Prompt 工程工具箱：6 个参考 + 一个 subagent","Awesome Prompt Engineering \u002F OpenAI Cookbook \u002F Prompt Architect 27 框架 + Claude Code prompt-engineer subagent。把 prompt 从猜变 checklist。","三个参考仓 + 两个框架合集 + 一个 Claude Code subagent，把 prompt 写作从猜测变成有 checklist 的手艺。","## 这个 pack 装了什么\n\n这个包凑齐了**六个高信号的 prompt 工程资产**，并配一个真的会用它们的 Claude Code subagent。组合是有意的：两个百科级参考，两个有立场的框架合集，两个能直接放进编辑器跑的工具。\n\n| # | 资产 | 类型 | 给你什么 |\n|---|---|---|---|\n| 1 | Awesome Prompt Engineering | 精选列表 | 论文 \u002F 课程 \u002F 库的索引 |\n| 2 | OpenAI Cookbook | 参考仓 | 200+ OpenAI API 的可跑示例 |\n| 3 | Prompt Architect — 27 框架 | 框架合集 | CRISPE \u002F RACE \u002F RICE \u002F RTF 等 27 个 |\n| 4 | Prompt Master | 框架合集 | 模式库带红队示例 |\n| 5 | prompt-engineer subagent | Claude Code agent | 按选定框架重写 prompt |\n| 6 | Prompt 脚手架 | 片段包 | 常见任务的可复用 system message |\n\n整个集合对操作顺序有立场：先读 awesome 列表摸清地图，挑一个适合任务的框架，然后用 subagent 跑你的草稿迭代。\n\n## 为什么叫\"工具箱\"而不是又一份长列表\n\n\"prompt engineering\"的搜索结果已经塌缩成每个页面同样的五条 tip。这个 pack 解决另一个问题：你已经会基本功之后，靠什么*变得更强*？\n\n答案是三件事拉扯：\n\n- **广度** —— 看 prompt 在不同领域 \u002F 模型 \u002F provider 怪癖下怎么变化。Awesome 列表和 Cookbook 覆盖\n- **结构** —— 选一个框架让 prompt 可审计 \u002F 可对比 \u002F 可复用。27 框架合集和 Prompt Master 覆盖\n- **迭代** —— 从第 N 版到 N+1 版要快，要有理由。Claude Code subagent 覆盖\n\n三件同时拥有才会复利。只占一两件，会让你接下来几个月反复踩同样的坑。\n\n## 一条命令装齐\n\n```bash\n# 装整个 pack\ntokrepo install pack\u002Fprompt-engineering-toolkit\n\n# 或者只装 subagent\ntokrepo install prompt-engineer\n```\n\nSubagent 在 Claude Code 里收到 `@prompt-engineer rewrite this prompt for clarity and falsifiability` 之类请求时激活。它默认用 CRISPE，输出 diff 加 rationale。框架合集存在 `.claude\u002Fskills\u002Fprompt-engineering\u002F` 下，任何会话都可以引用。\n\n## 常见坑\n\n- **把框架当圣经**：CRISPE \u002F RACE \u002F RTF 是脚手架，不是法律。Subagent 按任务挑一个；觉得选错了就用 `--framework=\u003Cname>` 强制覆盖，不要跟输出较劲\n- **跳过评测**：感觉更好的改写可能在你真实测试集上更差。配合 **LLM 评测 & 护栏**（Promptfoo \u002F DeepEval），让每次改动都有量化 delta\n- **provider 漂移**：OpenAI Cookbook 示例假设 OpenAI API。Claude \u002F Gemini 等价物在细节上不同（system message 处理 \u002F 工具 schema）。移植时先看 provider 官方 prompting 文档\n- **over-prompting**：长 prompt 藏 bug。一条 prompt 超过 ~400 token 时，把部分拆到工具定义或检索调用，不要全塞进 system message\n- **不进版本管理**：prompt 就是代码。commit \u002F diff \u002F code review 一样不少。Subagent 输出 diff 就是为了让这个生命周期跑得顺\n\n## 常见误解\n\n- *\"模型变聪明了，prompt engineering 死了。\"* 反过来 —— 越好的模型越奖励结构化 prompt，因为它能可靠遵循更多约束。死掉的是*奇技淫巧*（越狱咒语 \u002F 魔法词）。结构化 prompt 比以前更值钱\n- *\"用了 LangChain 这种框架就不需要这个。\"* 框架是组合 prompt，不是写 prompt。一条 LangChain chain 里的 system message 和 tool description 还是你要写的 prompt\n- *\"OpenAI Cookbook 只对 OpenAI 有用。\"* 模式（函数调用 \u002F 结构化输出 \u002F 评估器）能干净地移到 Claude \u002F Gemini。绑定不同，方法相同",[76,79,82,85,88],{"q":77,"a":78},"这个 pack 免费吗？","免费。每个资产都开源 —— 五个 GitHub 仓加上 Anthropic 格式的 subagent。TokRepo 安装免费，不引入代理或 token。你只在真用 subagent 跑草稿时为 LLM API 调用付费，账单走你用的 provider（Claude \u002F OpenAI \u002F Gemini）。",{"q":80,"a":81},"跟用 ChatGPT 重写 prompt 有啥区别？","ChatGPT 可以重写，但它会隐式选个框架，不给你 rationale。prompt-engineer subagent 显式选框架，列出加了哪些约束，输出 unified diff，让你能审查改了啥、为啥改。改写就变得可审计，可以拒绝单个选择，而不是整团接受。",{"q":83,"a":84},"Claude Code \u002F Cursor 都能用吗？","Subagent 是 Claude Code 原生（一个 .claude\u002Fagents\u002F*.md 文件）。框架合集和参考仓是语言无关的 —— 装成 Markdown，任何 AI 编辑器都能读。Cursor 用户一般通过 @-mention 引用，Codex CLI 用户放进 AGENTS.md。Subagent 必须要 Claude Code 的 agent 调用语法。",{"q":86,"a":87},"跟手写 prompt 比有啥不同？","手写在你以前 prompt 过这个任务时挺好。这个 toolkit 在你从零起步、或 prompt 出问题但说不清哪儿不对时发光。框架给你词汇（角色 \u002F 上下文 \u002F 具体性 \u002F 示例）来命名缺什么。等你内化了可能就不用 subagent 了 —— 这是成功，不是失败。",{"q":89,"a":90},"运维坑？","团队最大的错是不把 prompt 进版本管理。Prompt 改动是代码改动，风险等级一样（回归 \u002F 漂移 \u002F 归因）。把改后的 prompt 当作重构的函数对待：PR \u002F review \u002F eval \u002F merge。Subagent 输出 diff 就是为了让这个工作流自然而非空想。",{"@context":92,"@type":93,"name":94,"description":95,"numberOfItems":96,"publisher":97},"https:\u002F\u002Fschema.org","CollectionPage","Prompt Engineering Toolkit","Six curated assets for prompt engineering — references, frameworks, and a Claude Code subagent.",6,{"@type":98,"name":99,"url":100},"Organization","TokRepo","https:\u002F\u002Ftokrepo.com",[102,106,110],{"url":103,"anchor":104,"reason":105},"\u002Fzh\u002Fpacks\u002Fclaude-code-subagents","Claude Code 子代理精选","里面有 prompt-engineer 专家 subagent",{"url":107,"anchor":108,"reason":109},"\u002Fzh\u002Fpacks\u002Fanthropic-builders","Anthropic 开发栈","官方 prompting 指南与 skill 格式",{"url":111,"anchor":112,"reason":113},"\u002Fzh\u002Fpacks\u002Fllm-eval-guardrails","LLM 评测 & 护栏","量化你改的 prompt 是不是真的更好",[115,119,123],{"claim":116,"source_name":117,"source_url":118},"Anthropic publishes a prompt engineering interactive tutorial and best-practices guide","Anthropic Prompt Engineering docs","https:\u002F\u002Fdocs.claude.com\u002Fen\u002Fdocs\u002Fbuild-with-claude\u002Fprompt-engineering\u002Foverview",{"claim":120,"source_name":121,"source_url":122},"OpenAI Cookbook is the canonical example collection for OpenAI API prompting","openai\u002Fopenai-cookbook","https:\u002F\u002Fgithub.com\u002Fopenai\u002Fopenai-cookbook",{"claim":124,"source_name":125,"source_url":126},"Awesome Prompt Engineering is a curated list of prompt resources","promptslab\u002FAwesome-Prompt-Engineering","https:\u002F\u002Fgithub.com\u002Fpromptslab\u002FAwesome-Prompt-Engineering",473,"2026-05-02T15:10:00Z"]