Back to Details

prompt-generator

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Prompt Generator

提示词生成器

Create high-quality, structured prompts using meta-prompting best practices: task decomposition, expert personas, iterative verification, and hallucination minimization.
运用元提示词最佳实践创建高质量、结构化的提示词：任务分解、专家角色设定、迭代验证以及幻觉最小化。

Workflow

工作流程

Phase 1: Gather Requirements

阶段1：收集需求

Ask the user (one at a time, maximum 3 questions):
  1. Goal: "What is the primary goal or role of the system you want to create?"
  2. Output: "What specific outputs do you expect? (format, length, style)"
  3. Accuracy: "How should it handle uncertainty? (disclaim, ask for sources, or best-effort)"
Skip questions when answers are obvious from context. Minimize friction.
逐个向用户提问（最多3个问题）：
  1. 目标："你想要创建的系统的核心目标或角色是什么？"
  2. 输出要求："你期望得到什么样的具体输出？（格式、长度、风格）"
  3. 准确性要求："该系统应如何处理不确定性？（声明免责、要求提供来源，还是尽力而为）"
若从上下文可明显得知答案，则跳过对应问题，尽可能减少用户操作成本。

Phase 2: Decompose (if complex)

阶段2：任务分解（若需求复杂）

For complex requests, break into subtasks and assign expert personas:
  • Expert Writer — for copywriting, narrative, tone
  • Expert Analyst — for data, logic, verification
  • Expert Python — for code generation, computation
  • Expert [Domain] — for specialized knowledge
Each expert gets complete, self-contained instructions (no shared memory between experts).
Use "fresh eyes" — never assign the same expert to both create AND validate.
针对复杂请求，将其拆分为子任务并分配专家角色：
  • Expert Writer（专业写手） — 负责文案撰写、叙事、语气风格
  • Expert Analyst（专业分析师） — 负责数据处理、逻辑分析、验证工作
  • Expert Python（Python专家） — 负责代码生成、计算相关任务
  • Expert [Domain]（领域专家） — 负责专业知识相关内容
每位专家将获得完整、独立的指令（专家之间无共享记忆）。 遵循“旁观者清”原则——绝不安排同一专家同时负责内容创建与验证工作。

Phase 3: Generate the Prompt

阶段3：生成提示词

Consolidate into a single, cohesive prompt. Include all applicable sections, omit sections not relevant to the use case:
undefined
将所有信息整合为一个连贯的提示词，包含所有适用部分，省略与当前场景无关的内容：
undefined

Role

Role

[Short, direct role definition. Emphasize verification and disclaimers for uncertainty.]
[简短、直接的角色定义。强调对不确定性的验证与免责声明。]

Context

Context

[User's task, goals, background. Summarize clarifications from user input.]
[用户的任务、目标、背景信息。总结从用户输入中得到的澄清内容。]

Instructions

Instructions

  1. [Stepwise approach, including how to verify data]
  2. [Expert assignments if needed]
  3. [How to handle uncertain or missing information]
  1. [分步执行方法，包括数据验证方式]
  2. [若有需要，分配专家角色]
  3. [如何处理不确定或缺失的信息]

Constraints

Constraints

[Limitations: style, length, references, disclaimers]
[限制条件：风格、长度、参考来源、免责声明]

Output Format

Output Format

[Exact structure of the final output — bullets, paragraphs, code blocks, etc.]
[最终输出的精确结构——项目符号、段落、代码块等]

Reasoning

Reasoning

[OPTIONAL — include only if the user wants chain-of-thought or rationale. Otherwise, omit to keep the prompt concise.]
[可选——仅当用户需要思维链或推理依据时包含。 否则省略，以保持提示词简洁。]

Examples

Examples

[OPTIONAL — include when user provides input/output pairs or when examples significantly improve output quality. Omit for straightforward tasks.]

**Section inclusion guide:**
- Role, Context, Instructions, Constraints, Output Format — **always include**
- Reasoning — include only for complex analytical or multi-step tasks
- Examples — include when output quality depends on seeing concrete patterns
[可选——当用户提供输入/输出示例对，或示例能显著提升输出质量时包含。简单任务可省略。]

**模块包含指南：**
- Role、Context、Instructions、Constraints、Output Format — **必须包含**
- Reasoning — 仅在处理复杂分析或多步骤任务时包含
- Examples — 当输出质量依赖于具体示例参考时包含

Phase 4: Verify and Deliver

阶段4：验证与交付

  • Self-review: check for ambiguous instructions, missing constraints, or sections that could cause hallucination
  • If experts were used, note their review
  • Present the final prompt, organized and easy to follow
  • Offer to iterate if the user wants adjustments
  • 自我审核：检查是否存在模糊指令、缺失的限制条件，或可能导致幻觉的模块
  • 若使用了专家角色，需注明其审核情况
  • 呈现最终的提示词，确保结构清晰、易于理解
  • 主动提出可根据用户需求进行迭代优化

Principles

原则

  • Decompose complex tasks into smaller subtasks
  • Fresh eyes — separate creation from validation
  • Never guess — disclaim uncertainty, ask for data
  • Concise — only ask clarifying questions when critical
  • Iterative — verify before delivering, offer refinement
  • Section-aware — include only relevant sections, omit what doesn't apply
  • 分解任务：将复杂任务拆分为更小的子任务
  • 旁观者清：将内容创建与验证工作分离
  • 绝不猜测：对不确定性进行免责声明，主动请求数据支持
  • 简洁性：仅在必要时提出澄清问题
  • 迭代优化：交付前进行验证，提供优化服务
  • 模块适配：仅包含相关模块，省略无关内容