做什么
使用框架将用户提示转换为优化提示(RTF、RISEN、思想链、RODES、密度链、RACE、RISE、STAR、SOAP、CLEAR、GROW)
技能库 智客分类:Agent 工作流 prompt-engineer
使用框架将用户提示转换为优化提示(RTF、RISEN、思想链、RODES、密度链、RACE、RISE、STAR、SOAP、CLEAR、GROW)
官方网址:skills.sh
先看中文介绍;官方 description 原文单独保留,不改写 SKILL.md。
使用框架将用户提示转换为优化提示(RTF、RISEN、思想链、RODES、密度链、RACE、RISE、STAR、SOAP、CLEAR、GROW)
当:
按 Agent Skills 渐进披露:启动时只加载 name 与 description(约 100 token);任务匹配后才读入整份 SKILL.md 正文;scripts/、references/、assets/ 仅在需要时再读。 本文件正文结构:Purpose、When to Use、Workflow、Step 1: Analyze Intent、Step 2: Ask Clarifying Questions (Conditional)、Step 3: Select Framework(s)。 其中含规范建议的小节:分步指令、输入输出示例。
文件分析:这是一份仅含 SKILL.md 的指令型技能,代理激活后整份正文进入上下文。
Transforms user prompts into optimized prompts using frameworks (RTF, RISEN, Chain of Thought, RODES, Chain of Density, RACE, RISE, STAR, SOAP, CLEAR, GROW)
PurposeWhen to UseWorkflowStep 1: Analyze IntentStep 2: Ask Clarifying Questions (Conditional)Step 3: Select Framework(s)Critical Rules**NEVER:****ALWAYS:****Example 2: Complex Clear Prompt → RODES + Chain of Thought (No Clarification)****Example 4: Code Debugging → Chain of Thought**Notes
来源分类:skills.sh agent-skill
nameprompt-engineerdescription具体调用语法与可用工具以目标 Agent 客户端为准。 查看调用机制说明 ↗
先选择目标 Agent 和安装范围,保留技能包的附属文件,安装后检查客户端能否发现该技能。
复制安装指令给支持 Agent Skills 的代理,确认其中的目标目录与客户端匹配。
把 Agent Skill「prompt-engineer」安装到我的项目:SKILL.md 原文与官方 description 见 https://zicq.com/zh/skills/skl-29f12f3fdc012f1d-%E5%BF%AB%E9%80%9F%E5%B7%A5%E7%A8%8B%E5%B8%88.html 请存为 .cursor/skills/prompt-engineer/SKILL.md 或 .claude/skills/prompt-engineer/SKILL.md,frontmatter 的 name 与 description 保持原样,不要改写。
需要 Node.js 与 npx。先查看仓库技能列表,确认实际名称。
npx skills add 'https://github.com/sickn33/agentic-awesome-skills' --list
npx skills add 'https://github.com/sickn33/agentic-awesome-skills' --skill 'prompt-engineer'
CLI 会交互选择目标 Agent,默认安装到项目;用户级安装使用 -g。先通过查看命令核对仓库内容,再用 npx skills list 检查已安装技能。
This skill transforms raw, unstructured user prompts into highly optimized prompts using established prompting frameworks. It analyzes user intent, identifies task complexity, and intelligently selects the most appropriate framework(s) to maximize Claude/ChatGPT output quality.
The skill operates in "magic mode" - it works silently behind the scenes, only interacting with users when clarification is critically needed. Users receive polished, ready-to-use prompts without technical explanations or framework jargon.
This is a universal skill that works in any terminal context, not limited to Obsidian vaults or specific project structures.
Invoke this skill when:
Objective: Understand what the user truly wants to accomplish.
Actions:
Detection Patterns:
Objective: Gather missing information only when it is critical to framework selection or prompt quality.
Trigger Conditions — ask only if:
Question Limits:
Example Clarifying Exchange:
User: "help me with AI"
Step 2 (triggered — task type ambiguous):
"To craft the best prompt, I need one quick clarification:
1. What do you want to do with AI — build something, learn about it, or use an AI tool for a task?"
Critical Rule: When in doubt, skip clarification and generate the best prompt with available context. Over-asking breaks the "magic mode" experience.
Objective: Map task characteristics to optimal prompting framework(s).
Framework Mapping Logic:
| Task Type | Recommended Framework(s) | Rationale | |-----------|-------------------------|-----------| | Role-based tasks (act as expert, consultant) | RTF (Role-Task-Format) | Clear role definition + task + output format | | Step-by-step reasoning (debugging, proof, logic) | Chain of Thought | Encourages explicit reasoning steps | | Structured projects (multi-phase, deliverables) | RISEN (Role, Instructions, Steps, End goal, Narrowing) | Comprehensive structure for complex work | | Complex design/analysis (systems, architecture) | RODES (Role, Objective, Details, Examples, Sense check) | Balances detail with validation | | Summarization (compress, synthesize) | Chain of Density | Iterative refinement to essential info | | Communication (reports, presentations, storytelling) | RACE (Role, Audience, Context, Expectation) | Audience-aware messaging | | Investigation/analysis (research, diagnosis) | RISE (Research, Investigate, Synthesize, Evaluate) | Systematic analytical approach | | Contextual situations (problem-solving with background) | STAR (Situation, Task, Action, Result) | Context-rich problem framing | | Documentation (medical, technical, records) | SOAP (Subjective, Objective, Assessment, Plan) | Structured information capture | | Goal-setting (OKRs, objectives, targets) | CLEAR (Collaborative, Limited, Emotional, Appreciable, Refinable) | Goal clarity and actionability | | Coaching/development (mentoring, growth) | GROW (Goal, Reality, Options, Will) | Developmental conversation structure |
Blending Strategy:
Selection Criteria:
Critical Rule: This selection happens silently - do not explain framework choice to user.
Role: You are a senior software architect. [RTF - Role]
Objective: Design a microservices architecture for [system]. [RODES - Objective]
Approach this step-by-step: [Chain of Thought]
Details: [RODES - Details]
Output Format: [RTF - Format] Provide architecture diagram description, service definitions, and migration roadmap.
Sense Check: [RODES - Sense check] Validate that services are loosely coupled, independently deployable, and aligned with business domains.
**4.5. Language Adaptation**
- If original prompt is in Portuguese, generate prompt in Portuguese
- If original prompt is in English, generate prompt in English
- If mixed, default to English (more universal for AI models)
**4.6. Quality Checks**
Before finalizing, verify:
- [ ] Prompt is self-contained (no external context needed)
- [ ] Task is specific and measurable
- [ ] Output format is clear
- [ ] No ambiguous language
- [ ] Appropriate level of detail for task complexity
## Critical Rules
### **NEVER:**
- ❌ Assume information that wasn't provided - ALWAYS ask if critical details are missing
- ❌ Explain which framework was selected or why (magic mode - keep it invisible)
- ❌ Generate generic, one-size-fits-all prompts - always customize to context
- ❌ Use technical jargon in the final prompt (unless user's domain is technical)
- ❌ Ask more than 3 clarifying questions (avoid user fatigue)
- ❌ Include meta-commentary in the output ("This prompt uses...", "Note that...")
- ❌ Present output without code block formatting
- ❌ Mix languages inconsistently (if user writes in PT, respond in PT)
### **ALWAYS:**
- ✅ Analyze intent before generating (Step 1 is mandatory)
- ✅ Ask clarifying questions if critical information is ambiguous (Step 2 conditional)
- ✅ Select framework(s) based on task type and complexity (Step 3 mapping)
- ✅ Blend multiple frameworks when it improves prompt quality
- ✅ Adapt prompt length to original input complexity (simple → short, complex → detailed)
- ✅ Include output format specification in generated prompts
- ✅ Present final prompt in clean Markdown code block
- ✅ Make prompts self-contained (no dependency on external context)
- ✅ Use examples in complex prompts to illustrate expected output
- ✅ Validate prompt completeness before presenting (quality checks in Step 4.6)
### **Example 2: Complex Clear Prompt → RODES + Chain of Thought (No Clarification)**
**User Input:**
copilot> create prompt for: I need to build a product recommendation system considering purchase history, user preferences, and market trends
**Skill Internal Process:**
1. **Analyze:** Clear intent, complex task (ML system design), technical domain
2. **Clarify:** Skip (sufficient information)
3. **Select:** RODES (complex design) + Chain of Thought (reasoning needed)
**Skill Output:**
````markdown
Role: You are a senior machine learning engineer specializing in recommendation systems.
Objective: Design a product recommendation system that combines user purchase history, preferences, and market trends.
Details:
Approach step-by-step:
Examples:
Sense Check: Validate that the proposed solution balances personalization with discovery, scales to production load, and incorporates real-time market trends effectively.
Output format: Technical design document with architecture diagrams, algorithm justification, and implementation roadmap.
User Input:
copilot> optimize prompt: my Python code isn't working, needs debugging
Skill Internal Process:
Skill Output:
```
You are an expert Python debugger.
Task: Debug the Python code provided by the user by analyzing it step-by-step.
Debugging approach:
1. Read and understand the code's intended purpose
2. Identify the error or unexpected behavior
3. Trace variable states at each execution step
4. Locate the root cause of the issue
5. Propose fix with explanation
6. Suggest preventive measures to avoid similar bugs
For each step, show your reasoning:
- What you're checking
- What you found
- Why it matters
Output format:
- **Issue identified:** [Describe the bug]
- **Root cause:** [Why it's happening]
- **Fix:** [Corrected code with comments]
- **Prevention:** [Best practices to avoid this in the future]
Include a working example to verify the fix.
```
This skill is platform-agnostic and works in any terminal context where GitHub Copilot CLI is available. It does not depend on:
The skill is entirely self-contained, operating purely on user input and framework knowledge.
Agent 工作流
创造有效技能指南。 当用户想创造出新的技能(或更新现有的技能),以专业知识,工作流程,或工具集成来扩展克洛德的能力时,应该使用这种技能.
Agent 工作流
使用ClawdHub CLI搜索,安装,更新并发布从taladhub.com的代理技能. 需要获取苍蝇上的新技能时使用,将安装的技能同步到最新版本或特定版本,或者发布 npm-instainddhub CLI 的新/更新的技能文件夹.
Agent 工作流
管弦乐团多代理团队,任务设定周期,交接协议,审查工作流程. 使用时间: (1)建立2+特派员队伍,具有不同专业,(2)确定任务路线和生命周期(收录框_ spec_建设_审查_完成),(3)在特派员之间制定交接协议,(4)建立审查和质量关口,(5)管理特派员之间的交流和文物共享.
Agent 工作流
Spec-first,TDD,子代理驱动的软件开发工作流程. 当:(1)构建任何新功能或应用——触发脑暴_计划_子代理执行回路,(2)调试出一个bug或测试失败——触发系统性的根起过程,(3)用户说"让我们构建","帮助我计划","我想添加X",或"这个被打破",(4)完成一个功…