Claude Code（Codex）中文提示词与英文提示词差异（RFC征求意见稿、ADR架构决策记录文档）Claude提示词、Codex提示词

Dontla

412人浏览 · 2026-05-22 01:57:06

Dontla · 2026-05-22 01:57:06 发布

对 Codex / Claude Code / GPT 系 Coding Agent 来说：

英文 System Design 文档通常效果更稳定。

原因不是“中文不行”，而是：

一、训练数据分布问题

这些 coding agent 的高质量训练数据，大部分来自：

GitHub
RFC (Request for Comments 一种技术文档标准格式，最初用于互联网标准制定)
ADR(Architecture Decision Record 架构决策记录文档)
OpenAPI
Architecture Docs
Tech Specs
PRD
Issues
StackOverflow

而这些：

绝大部分是英文。

所以 agent 对这种英文结构：

Responsibilities:
Outputs:
Constraints:
Success Criteria:

理解非常强。

它会自动联想到：

模块边界
接口设计
工程分层
workflow
dependency

二、中文不会“看不懂”

实际上：

现在 GPT-5 / Claude 4 已经完全能理解中文工程文档。

所以：

中文完全能做项目。

尤其：

产品描述
PRD
workflow
prompt
用户需求

都没问题。

但问题在于：

中文工程术语容易“漂移”

比如：

你写：

机会发现

Agent 可能理解成：

recommendation
lead generation
trend analysis
opportunity mining

但英文：

Opportunity Discovery

含义会稳定很多。

三、最容易出问题的是“抽象描述”

例如中文：

后面做一些评估

LLM 很容易自由发挥。

但英文：

Evaluate technical feasibility and market viability

边界会清晰很多。

四、真正推荐的方法

不是“全英文”。

而是：

中文思考 + 英文工程定义

这是现在很多 AI 工程团队真实做法。

例如：

你自己用中文整理：

我希望系统从社媒发现用户痛点

然后转成：

The system discovers user pain points from social discussions.

再比如：

中文：

后续进行价值评估

转：

Evaluate market value, technical feasibility, and competitive landscape.

五、最佳实践（非常推荐）

你以后所有 AI 工程项目：

建议采用：

1. 中文思考

用于：

brainstorming
产品构思
workflow 推演
agent 设计

因为：

中文对你认知负担最低。

2. 英文工程文档

用于：

README
Architecture
PRD
Codex instructions
Prompt templates
Agent contracts
API schema

因为：

LLM 对英文工程语义最稳定。

六、尤其是这些东西，强烈建议英文

以下内容最好永远英文：

Agent 名称

用：

Collector Agent
Ranking Agent
Evaluation Agent

不要：

收集代理
排序代理

因为英文 agent naming 更稳定。

状态流转

pending
running
completed
failed
retrying

不要中文。

Artifact 类型

Opportunity Report
PRD
Architecture Doc
Task Breakdown

Prompt

Prompt 强烈建议英文。

因为：

英文 prompt：

更短
token 更少
语义边界更稳定

七、什么时候中文反而更好

如果：

你的项目：

面向中文用户
分析中文社媒（尤其 Rednote）
提取中文语义
中文情绪分析

那：

数据层必须支持中文

包括：

中文 embedding
中文 chunking
中文 normalization

因为：

中文互联网语义和英文互联网差异极大。

八、你现在最推荐的方式

你现在应该：

用英文写：

项目定义
agent contract
architecture
workflow
artifact schema

用中文：

思考
产品推演
灵感记录

这是目前 AI 工程效率最高的方式。

提示词示例

# Project Vision

This project is a multi-agent opportunity discovery and evaluation system.

The system continuously collects discussions, complaints, feature requests, and trend signals from multiple social platforms, then uses a series of AI agents to:

1. Discover emerging user needs
2. Cluster and rank opportunities
3. Evaluate market and technical feasibility
4. Generate structured business and engineering artifacts

The long-term goal is to build an AI-native "opportunity intelligence pipeline" that transforms raw internet discussions into actionable product and engineering outputs.

---

# Core Concept

This is NOT just a chatbot or autonomous agent demo.

The system is designed as an Artifact Pipeline:

Social Signals
    ↓
Need Discovery
    ↓
Clustering & Ranking
    ↓
Evaluation
    ↓
Planning
    ↓
Artifact Generation

The final output of the system is structured artifacts, not conversations.

---

# Information Sources

The MVP should support:

- Reddit

Future versions may support:

- Hacker News
- GitHub Issues
- Twitter/X
- Product Hunt
- Discord
- Telegram
- YouTube comments
- App Store reviews
- Rednote (Xiaohongshu)

---

# MVP Scope

The MVP should remain intentionally small and focused.

DO NOT over-engineer the system.

Avoid:
- distributed systems
- microservices
- autonomous self-improving agents
- complex memory systems
- multi-region infrastructure
- advanced scheduling systems

The MVP goal is to validate the workflow and artifact pipeline.

---

# MVP Architecture

The MVP should contain 5 core agents.

## 1. Collector Agent

Responsibilities:
- collect social posts
- normalize data
- clean content
- deduplicate records

Outputs:
- raw signal dataset

---

## 2. Signal Ranking Agent

Responsibilities:
- embedding generation
- clustering
- trend detection
- pain point extraction
- signal ranking

Outputs:
- ranked opportunity candidates

---

## 3. Evaluation Agent

Responsibilities:
- market analysis
- technical feasibility
- competitor analysis
- risk analysis
- confidence scoring

Outputs:
- evaluated opportunities

---

## 4. Planner Agent

Responsibilities:
- PRD generation
- architecture planning
- task breakdown generation
- implementation suggestions

Outputs:
- structured engineering plans

---

## 5. Document Agent

Responsibilities:
- markdown generation
- JSON export
- report formatting
- artifact persistence

Outputs:
- final artifacts

---

# MVP Artifacts

The MVP should generate 4 primary artifacts.

## 1. Opportunity Report

Contains:
- identified problem
- supporting evidence
- user pain analysis
- market signals
- confidence score

---

## 2. PRD

Contains:
- product goals
- user stories
- functional requirements
- non-functional requirements

---

## 3. Architecture Document

Contains:
- system components
- workflow design
- infrastructure suggestions
- storage design

---

## 4. Task Breakdown

Contains:
- implementation tasks
- priorities
- estimated complexity
- milestone suggestions

---

# Suggested Tech Stack (MVP)

Backend:
- Python
- FastAPI

AI Orchestration:
- LangGraph

LLM:
- OpenAI / Anthropic APIs

Vector Storage:
- pgvector or Chroma

Database:
- PostgreSQL

Async Tasks:
- asyncio initially

Frontend:
- simple dashboard
- minimal UI

Observability:
- basic logging only

Deployment:
- Docker Compose

---

# Recommended Repository Structure

monorepo preferred

Example:

/apps
    /api
    /worker
    /frontend

/packages
    /agents
    /prompts
    /shared

/docs

/artifacts

---

# Production Vision

Future production versions may include:

## Workflow Infrastructure
- Temporal
- durable execution
- retries
- checkpointing

## Advanced Agent Runtime
- long-running workflows
- human-in-the-loop
- memory systems
- agent versioning

## Observability
- Langfuse
- OpenTelemetry
- tracing
- token/cost tracking

## Scalability
- Kafka
- Redis streams
- distributed workers

## Advanced Evaluation
- evaluator agents
- reinforcement loops
- historical opportunity tracking

## Artifact Expansion
Additional artifacts may include:
- Jira tickets
- GitHub issues
- startup briefs
- investor reports
- architecture diagrams
- sprint plans

---

# Engineering Principles

The system should prioritize:

1. Simplicity over complexity
2. Deterministic workflows over autonomy
3. Structured artifacts over chat responses
4. Clear state transitions
5. Human-readable outputs
6. Observable workflows
7. Modular agents

---

# Important Constraints

- Keep the MVP implementation small
- Avoid premature optimization
- Prefer explicit workflows over "magic"
- Build for clarity and maintainability
- Every agent must have clear inputs and outputs
- Every workflow stage should produce structured artifacts

---

# Success Criteria for MVP

The MVP is considered successful if it can:

1. Collect real social signals
2. Detect at least one meaningful opportunity
3. Generate useful structured artifacts
4. Demonstrate a complete end-to-end workflow
5. Produce outputs that humans can review and act on

The goal is NOT full autonomy.

The goal is a reliable AI-assisted opportunity pipeline.