decebals/claude-code-java

What it solves

It provides a standardized set of "skills"—structured domain knowledge and workflows—to help AI agents provide consistent, high-quality assistance for Java and Maven projects. It solves the problem of inconsistent AI behavior by giving agents specific guidelines for tasks like code reviews, architecture decisions, and dependency audits, while reducing token usage through optimized patterns.

How it works

The project consists of a collection of markdown files that follow the open Agent Skills specification. These files are symlinked or copied into a Java project's .claude/skills directory. When using an agent like Claude Code, the agent automatically loads the relevant skill based on the task description. The project includes setup scripts to automate the configuration of these skills, project templates, and MCP server configurations.

Who it’s for

Java developers who want to integrate AI agents into their development workflow for tasks such as code reviews, testing, commit generation, and architectural refactoring.

Highlights

• 18 Specialized Skills: Covers workflows (git commits, changelogs), code quality (security audits, performance smells), architecture (SOLID, design patterns), and frameworks (Spring Boot, JPA).
• Spec Compliant: Adheres to the Agent Skills specification and includes CI validation to ensure skills are correctly formatted.
• Routing Validation: Includes a testing suite (eval-routing.sh) to ensure the AI agent selects the correct skill for a given prompt.
• Integrated Ecosystem: Works with Claude Code and can be used as a source of truth for automated PR reviews via GitHub Actions.