English | 简体中文
A Claude Code Skill that generates a data admin backend from a single sentence.
Powered by the erupt low-code framework: say "build me a library admin backend" and get a complete, runnable admin system — login page, CRUD, search, import/export, and permission management out of the box.
No dev environment required: with no system JDK, the skill downloads OpenJDK 25 on demand (Adoptium official source, with a Tsinghua mirror fallback that's fast in China; cached in ~/.erupt-skill), Maven is prepared automatically, project dependencies go through the Aliyun mirror, and an embedded H2 file database means no MySQL install. The repo itself stays ~1 MB, so cloning is instant. Non-developers can build a data admin backend for any domain with one sentence.
This skill follows the open Agent Skills standard (SKILL.md), supported by most mainstream AI coding tools. Clone the repo into the skills directory of your tool:
| Tool | Skills directory | Notes |
|---|---|---|
| Claude Code | ~/.claude/skills/ (personal) or .claude/skills/ (project) |
Native support, auto-activates by description |
| Codex CLI | ~/.codex/skills/ (personal) or .codex/skills/ (project) |
Invoke with $erupt-admin ... or auto-activation |
| Cursor | .cursor/skills/ or .agents/skills/ (project) |
See Cursor Agent Skills docs |
| Trae | .agents/skills/ (project) |
Supports the Agent Skills open standard |
| CodeBuddy | .agents/skills/ (project) |
Supports the Agent Skills open standard |
Example (Claude Code):
git clone https://github.com/plinian/erupt-skill.git ~/.claude/skills/erupt-adminFor any tool without native skill support, just clone this repo anywhere and tell the agent: "Read SKILL.md in this repo and follow its workflow to build me a XXX admin backend."
Just say in Claude Code:
Build a library admin backend
Create a customer order management system with status filtering
Add a borrow-record module to the backend you just built
Then open http://localhost:8080 and log in with erupt / erupt.
SKILL.md # main skill workflow
reference/annotations.md # entity generation spec (decision tables + common pitfalls)
reference/erupt-*.md # official erupt reference docs (annotations, i18n, query, Cube, API, TPL)
template/ # runnable erupt project template (latest erupt release + Spring Boot 3.5 + H2)
└─ resources/public/ # app.js (eruptSiteConfig: title/logo/theme/hooks) · app.css (styles)
scripts/setup-env.sh # prepare JDK + Maven (system JDK → cached → download JDK: Adoptium official, Tsinghua mirror fallback)
scripts/compile.sh # compile-only check (catch syntax errors after code changes)
scripts/run.sh # one-command build and start
scripts/build-m2-seed.sh # optional: build a local dependency seed into vendor/m2 for fully-offline first builds
- macOS / Linux / Windows (on Windows run inside Git Bash, which Claude Code ships with)
- Network access (first run downloads Maven and project dependencies; with no system JDK it also downloads OpenJDK 25 once, then caches it)