什么是 软件工程师？

软件工程师技能旨在通过生成符合高级工程师标准的编码，提升 AI 驱动的开发水平。它优先采用“先读后写”的方法，确保每一行新代码都尊重项目现有的风格、命名规范和依赖版本。通过将此技能集成到您的 Openclaw Skills 库中，您可以让您的智能体做出明智的架构决策，平衡交付速度与长期可维护性。

该技能通过在处理程序、服务和存储库之间强制执行严格的边界，超越了简单的代码生成。它将错误视为一等公民，摒弃通用的 catch 块，转向类型化、上下文感知的错误处理。结果是代码读起来像散文一样——简单、可预测且随时可以进行拉取请求（PR）审查。

下载入口:https://github.com/openclaw/skills/tree/main/skills/ivangdavila/software-engineer

安装与下载

1. ClawHub CLI

从源直接安装技能的最快方式。

2. 手动安装

将技能文件夹复制到以下位置之一

优先级：工作区 > 本地 > 内置

3. 提示词安装

将此提示词复制到 OpenClaw 即可自动安装。

软件工程师 应用场景

• 在保持严格架构层的同时，从零开始构建新功能。
• 将复杂的遗留函数重构为模块化、可测试的组件。
• 在关键业务路径中实现稳健的错误处理和日志记录。
• 记录架构折中方案，确保技术债得到跟踪和管理。
• 自动化代码审查，以识别反模式、幻数和过度工程。

1. 该技能首先扫描当前代码库，以映射出既有的模式、库和项目结构。
2. 它根据 YAGNI（你不需要它）原则评估特定需求，以避免过度工程。
3. 遵循分层的关注点分离编写代码，确保业务逻辑与基础设施细节隔离。
4. 在逻辑中注入稳健的错误处理，提供上下文并区分致命故障与可恢复故障。
5. 应用最终的 PR 就绪检查清单，确保函数简洁、边缘情况得到处理且删除了死代码。

软件工程师 配置指南

要在您的环境中激活此技能，请确保已在智能体配置中注册。Openclaw Skills 旨在与您现有的开发工具无缝协作。

# 通过 CLI 安装
clawdbot install software-engineer

安装完成后，为智能体提供对 patterns.md 或 architecture.md 文件的访问权限，以便根据您的特定组织标准进一步定制其输出。

软件工程师 数据架构与分类体系

软件工程师技能通过以下元数据和文档文件，利用结构化方法实现项目情报：

数据分层处理，从处理程序（验证）到服务（逻辑）再到存储库（数据访问），确保没有任何一层成为瓶颈或紧耦合的源头。

name: Software Engineer
slug: software-engineer
version: 1.0.0
homepage: https://clawic.com/skills/software-engineer
description: Write production-ready code with clean architecture, proper error handling, and pragmatic trade-offs between shipping fast and building right.
metadata: {"clawdbot":{"emoji":"?????","requires":{"bins":[]},"os":["linux","darwin","win32"]}}

When to Use

Agent needs to write, review, or refactor code. Handles implementation decisions, architecture trade-offs, and code quality across any language or framework.

Quick Reference

Core Rules

1. Read Before Write

• Check existing code style, patterns, and conventions before writing new code
• Respect the current stack — never swap libraries without explicit request
• Match naming conventions, formatting, and project structure already in place

2. Code That Compiles

Every code block must:

• Have correct imports for the actual library versions in use
• Use APIs that exist in the project's dependency versions
• Pass basic syntax checks — no placeholder // TODO: implement

3. Minimal First

• Solve the specific problem, not hypothetical future problems
• One abstraction when you have three concrete cases, not before
• Features that might be needed → skip. Features that are needed → implement

4. Errors as First-Class Citizens

? catch (e) {}
? catch (e) { console.log(e) }
? catch (e) { logger.error('context', { error: e, input }); throw new DomainError(...) }

• Typed errors over generic strings
• Include context: what operation failed, with what input
• Distinguish recoverable vs fatal errors

5. Boundaries and Separation

6. Explicit Trade-offs

When making architectural choices, state:

• What you chose and why
• What you traded away
• When to revisit the decision

Example: "Using SQLite for simplicity. Trade-off: no concurrent writes. Revisit if >1 write/sec needed."

7. PR-Ready Code

Before delivering any code:

• No dead code, commented blocks, or debug statements
• Functions under 30 lines
• No magic numbers — use named constants
• Early returns over nested conditionals
• Edge cases handled: null, empty, error states

Code Quality Signals

Senior code reads like prose:

• Names explain "what" and "why", not "how"
• A junior understands it in 30 seconds
• No cleverness that requires comments to explain

The best code is boring:

• Predictable patterns
• Standard library over dependencies when reasonable
• Explicit over implicit

Common Traps

Pragmatic Shipping

Critical paths (do it right):

• Authentication, authorization
• Payment processing
• Data integrity, migrations
• Secrets management

Experimental paths (ship fast, iterate):

• UI/UX features
• Admin panels
• Analytics
• Anything unvalidated with users

Test for critical path: "Can this wake me at 3am or lose money?"

Security & Privacy

This skill does NOT:

• Store any data externally
• Make network requests
• Access files outside the current project

All code suggestions are generated in context of the conversation.