opencode

Opencode automates website creation, project analysis, script execution, file management, server preview, and code generation tasks efficiently.

Security Scan

The skill's files and runtime instructions are coherent with its stated purpose (website/project automation and local file ops); it does perform broad filesystem and process operations so review before running on sensitive hosts.

Purpose & Capability

Name/description (website creation, project analysis, script execution, file management, preview servers) line up with the included SKILL.md, README, and the shell script. The examples and commands (opencode CLI, cp, tar, python3 -m http.server, find, git, npm/pip list) are expected for this functionality.

Instruction Scope

SKILL.md and scripts instruct the agent to run many shell commands that read and modify arbitrary filesystem paths (find, cp -r, tar, rm -rf, cat, git status) and start local servers. This is consistent with the skill's purpose but grants broad file and process access; the instructions also reference user-specific paths (e.g., /home/linshui/.openclaw/workspace) and opencode config/log locations (~/.config/opencode, ~/.local/share/opencode). Review and limit target paths when running.

Install Mechanism

This is instruction-only with baked-in helper shell script; there is no install spec, no remote downloads, and no package installation performed by the skill package itself. Risk from install mechanism is low.

Credentials

The skill declares no required environment variables or credentials (which is reasonable), but SKILL.md and the scripts reference configuration files, cache paths, and optional OPENCODE_* environment variables (e.g., OPENCODE_CACHE, OPENCODE_MAX_MEMORY, OPENCODE_SERVER_PASSWORD). The script also deletes cache (~/.local/share/opencode/cache/*) and temp files (/tmp/opencode_*). Those actions can affect local state and could remove user data if run blindly. No external credentials are requested, which is proportional, but the skill will read and modify local config/cache/log files.

Persistence & Privilege

Flags: always is false and default autonomy is enabled (normal). The skill does not request system-wide persistent privileges or claim it will modify other skills. It can start background servers and run continuous monitoring loops (the script's monitor function), so exercise caution about long-running or network-exposed services.

Assessment

This package appears coherent with its stated purpose, but it runs real shell commands that read, modify, compress, delete, and serve files. Before installing or invoking: 1) Review the included scripts (scripts/opencode-automation.sh) and SKILL.md and ensure target paths are safe; do not run as root. 2) Run first in a sandbox or throwaway VM if you want to verify behavior. 3) Be cautious with commands that use rm -rf, cp -r, or operate on ~, /tmp, or system paths—back up important data. 4) Confirm the opencode binary you call is the legitimate one (check opencode --version and its origin) because the skill delegates actions to that CLI which may connect to the network. 5) If you need to limit risk, restrict the agent to a specific workspace directory and firewall/block any unwanted network exposure for local servers.

Opencode 自动化技能

关于 Opencode 技能

Opencode 是一个强大的 AI 工作流引擎，可以帮助您：

• 🚀 快速创建网站和网页 - 从概念到成品，一键生成
• 📊 分析项目 - 查看项目结构、文件内容、统计信息
• 🛠️ 运行脚本和任务 - 执行自动化任务
• 📁 文件管理 - 创建、复制、移动、删除文件
• 🌐 Web 服务器管理 - 启动本地服务器预览项目
• 📝 代码生成 - 生成各种类型的代码文件
• 🔍 项目分析 - 分析项目状态、依赖关系

使用场景

当您需要：

1. 快速创建网站/网页 - "帮我创建一个电商网站"、"生成一个个人博客"、"设计一个公司官网"

2. 分析现有项目 - "查看这个项目的结构"、"分析这个文件夹的内容"、"统计项目文件"

3. 运行自动化任务 - "执行这个项目的所有测试"、"运行 linting 检查"

4. 文件操作 - "复制这些文件到另一个位置"、"压缩这个项目"、"提取项目中的图片"

5. 启动预览服务器 - "我想在浏览器中查看这个网站"、"启动本地服务器"

6. 代码生成 - "生成一个 REST API"、"创建数据库模型"、"编写单元测试"

7. 项目分析 - "这个项目使用了哪些技术？"、"分析项目的依赖关系"

核心功能

1. 启动 opencode 服务器

# 启动 Web 界面
opencode web --port 4096 --hostname 127.0.0.1

# 启动后台服务
opencode serve --port 4096

# 启动 ACP 服务器
opencode acp --port 4096

2. 运行 opencode 任务

# 直接运行任务
opencode run "任务描述..."

# 运行并等待完成
opencode run --wait "任务描述..."

# 运行并查看输出
opencode run --output "任务描述..."

3. 连接到 opencode 服务器

# 连接到正在运行的服务器
opencode attach http://127.0.0.1:4096

# 使用不同的端口
opencode attach http://localhost:8080

4. 分析项目

# 查看项目结构
tree -L 2 /path/to/project

# 统计项目大小
du -sh /path/to/project

# 查看文件列表
find /path/to/project -type f -name "*.py" | wc -l

# 查看 Git 状态
cd /path/to/project && git status

# 查看依赖
cd /path/to/project && npm list  # 或 yarn list / pip list

5. 文件操作

# 复制文件
cp -r /source /destination

# 压缩项目
tar -czf /path/to/project.tar.gz /path/to/project

# 创建压缩包
cd /path/to/project && zip -r ../project.zip .

# 提取压缩包
tar -xzf /path/to/archive.tar.gz

# 查看文件内容
cat /path/to/file
head -100 /path/to/file
tail -50 /path/to/file

# 搜索文件
grep -r "pattern" /path/to/project

# 查找特定类型的文件
find /path/to/project -type f -name "*.html"

6. 启动预览服务器

# Python HTTP 服务器
python3 -m http.server 8000 -d /path/to/project

# 使用不同的端口
python3 -m http.server 3000 -d /path/to/project

# Node.js 静态文件服务器
npx http-server -p 8080 -c-1 /path/to/project

# 使用 Python 的内置模块
python3 -m http.server 8000 --bind 127.0.0.1

工作流程

创建网站/网页的工作流

1. 定义需求 → 2. 启动 opencode → 3. 输入任务 → 4. 等待生成 → 5. 查看结果

示例：

opencode run "创建一个科技公司官网，包含导航栏、Hero 区域、公司简介、核心服务、产品展示、技术优势、客户案例、团队介绍、新闻动态、联系我们。科技感配色，响应式设计。保存到 /home/linshui/.openclaw/workspace/project-name"

分析项目的工作流

1. 选择项目 → 2. 查看结构 → 3. 分析内容 → 4. 生成报告

示例：

# 查看项目结构
find /path/to/project -type f | head -50

# 分析技术栈
# 查看 package.json / requirements.txt / go.mod 等

# 统计代码行数
wc -l /path/to/project/**/*.py

预览项目的工作流

1. 选择项目 → 2. 启动服务器 → 3. 打开浏览器 → 4. 查看效果

示例：

# 启动服务器
python3 -m http.server 8000 -d /path/to/project

# 在浏览器中打开
# http://127.0.0.1:8000

常用命令速查表

命令	说明	示例
opencode run	运行任务	opencode run "创建网站"
opencode web	启动 Web 界面	opencode web --port 4096
opencode serve	启动后台服务	opencode serve --port 4096
opencode attach	连接到服务器	opencode attach http://localhost:4096
opencode acp	启动 ACP 服务器	opencode acp --port 4096
opencode models	查看模型列表	opencode models
opencode stats	查看使用统计	opencode stats --days 7
opencode export	导出会话	opencode export <session_id>
opencode import	导入会话	opencode import file.json

注意事项

1. 端口占用 - 确保使用的端口没有被其他程序占用
2. 权限 - 某些操作可能需要 sudo 权限
3. 网络 - opencode 可能需要网络连接来下载模型
4. 存储 - 确保有足够的磁盘空间用于缓存和临时文件

系统集成

该技能可以与 OpenClaw 的其他技能配合使用：

• web_search - 搜索 opencode 相关资源
• coding-agent - 生成 opencode 脚本
• browser - 在浏览器中查看 opencode 生成的内容
• exec - 执行 opencode 命令

---

版本: 1.0.0 最后更新: 2026-03-17 作者: linshui