Deploy Agent

Automatically detects project type, generates Dockerfile if needed, builds and deploys the app in Docker with smart port selection and health checks.

Deploy-Agent Skill

Local deployment agent that auto-detects project types and deploys them via Docker. Default workflow: detect → Dockerfile generate → build → deploy.

Architecture

text

skills/deploy-agent/
├── SKILL.md                          ← This file (instructions for OpenClaw agent)
├── scripts/
│   ├── deploy-agent.sh               ← Main orchestrator (all-in-one CLI)
│   ├── bootstrap-docker.sh           ← Install Docker Engine + Compose
│   ├── detect-project.sh             ← Auto-detect project type
│   ├── generate-dockerfile.sh        ← Generate Dockerfile dynamically
│   └── templates/
│       ├── Dockerfile.node           ← Node.js template
│       ├── Dockerfile.python         ← Python/Django template
│       ├── Dockerfile.go             ← Go template
│       ├── Dockerfile.static         ← Static Nginx template
│       ├── Dockerfile.java           ← Java Maven template
│       └── entrypoint.django.sh      ← Django entrypoint (migration + run)

Capabilities

Feature	Description
Auto-detect	Identifies project type from file structure (Node.js, Python, Go, Rust, static HTML, Java, etc.)
Docker auto-gen	Generates Dockerfile if none exists
Compose native	Detects docker-compose.yml and deploys with compose
Smart ports	Auto-detects project port and finds next available
Health checks	Built-in HEALTHCHECK in every generated Dockerfile
Restart policy	Containers set to --restart unless-stopped
Managed tracking	All containers labeled with deploy-agent.managed=true

Supported Project Types

• Node.js — npm, yarn, pnpm; frameworks: Express, Fastify, NestJS, Next.js, Nuxt, SvelteKit, Astro, Vite
• Python — pip, pipenv, poetry; frameworks: Django, Flask, FastAPI
• Go — go.mod detection, multi-stage build
• Rust — Cargo.toml, multi-stage build
• Static HTML — nginx-alpine serve
• Java — Maven (pom.xml), multi-stage
• Docker Compose — native support
• Ruby/Rails, PHP/Laravel, Deno — basic detection (Dockerfile under development)

Usage (CLI)

bash

# 1. Install Docker (first time only)
bash scripts/deploy-agent.sh install-docker

# 2. Detect a project
bash scripts/deploy-agent.sh detect ~/my-project

# 3. Deploy (auto-detect, generate Dockerfile, build, run)
bash scripts/deploy-agent.sh deploy ~/my-project

# 4. Deploy with specific port
bash scripts/deploy-agent.sh deploy ~/my-project 8080

# 5. List managed containers
bash scripts/deploy-agent.sh status

# 6. Stop a container
bash scripts/deploy-agent.sh stop my-project

# 7. View logs
bash scripts/deploy-agent.sh logs my-project

Usage (OpenClaw Agent — this file)

When the user says "帮我部署这个项目" or provides a project path:

Step 1: Resolve project path

The project path may come as:

• A local path (e.g. ~/projects/my-app)
• A git URL (e.g. https://github.com/user/repo.git)
• A zip/tar archive (extract first)
• An attached file/folder (use the provided path)

Step 2: Check Docker availability

bash

# If Docker not available, install
bash <skill_dir>/scripts/bootstrap-docker.sh

Step 3: Detect project

Read detection output, confirm with user if ambiguous:

bash

bash <skill_dir>/scripts/detect-project.sh <project_dir>

Step 4: Deploy

bash

bash <skill_dir>/scripts/deploy-agent.sh deploy <project_dir>

Step 5: Report results

Tell the user:

• Project type detected
• Port the service is running on
• Container name
• How to access (http://host:port)
• How to stop / view logs

Uploaded projects

When user uploads project files:

1. Determine where files landed (check /tmp/, ~/Downloads/, workspace paths)
2. If it's a single archive (zip/tar.gz), extract it
3. Run detection on the extracted/landed directory
4. If it has Dockerfile → build+deploy
5. If docker-compose → compose up
6. If detectable (Node/Python/Go/etc) → generate Dockerfile, build, deploy
7. If unknown → tell user what was found and ask for guidance

Edge Cases

• Port conflict: If the port is already in use, deploy-agent auto-increments to find a free port
• Permission denied: User may need sudo usermod -aG docker $USER or newgrp docker
• Large projects: Node.js node_modules already present → COPY will be large. Suggest .dockerignore generation
• PostgreSQL/Redis dependencies: If the project needs databases, generate a docker-compose.yml instead

Script Reference

deploy-agent.sh

Main CLI: deploy-agent.sh <command> [args] Commands: install-docker, detect, generate, build, deploy, status, stop, logs

bootstrap-docker.sh

Idempotent Docker Engine + Compose Plugin installation for Ubuntu/Debian. Safe to run multiple times.

detect-project.sh

Scans project directory structure, outputs TYPE, FRAMEWORK, BUILD_CMD, RUN_CMD, PORT. Exit codes: 0 = detected, 1 = unknown.

generate-dockerfile.sh

Reads detection output, selects template, writes Dockerfile. Skips if Dockerfile/docker-compose.yml already exists.

Container Management

All deploy-agent containers are labeled:

• deploy-agent.managed=true
• deploy-agent.source=<absolute-source-dir>

To clean up all:

bash

docker rm -f $(docker ps -aq --filter "label=deploy-agent.managed=true")

⚖️ 权限声明

权限	范围	用途	说明
执行	sudo	Docker 安装、用户组管理	bootstrap-docker.sh 安装 Docker
执行	docker	容器管理	构建、运行、停止容器
文件系统	读取	项目目录	检测项目类型、读取构建配置
文件系统	写入	项目目录	生成 Dockerfile
网络	出站	Docker Hub / 镜像源	拉取基础镜像
网络	端口绑定	宿主机端口	容器端口映射，检测可用端口

⚠️ 除非使用 install-docker 命令，否则不需要 sudo 权限。Docker 操作建议通过 docker 用户组运行。