Summary

This repository ports eleven advanced Claude Code AI skills, including code review, security auditing, and feature development, to the OpenCode platform. It enables OpenCode users to leverage sophisticated multi-agent AI workflows previously exclusive to Anthropic's ecosystem, enhancing developer productivity and code quality.

How It Works

The project translates Claude Code's multi-agent workflows and skills into OpenCode's native SKILL.md format. These skills are exposed both through OpenCode's standard skill discovery mechanism and as physical slash command files (commands/*.md) that directly inline the full skill instructions. This dual-surface approach ensures reliable invocation and avoids potential caching issues, while enhanced review skills are designed to be more directive for potentially smaller models.

Quick Start & Requirements

Installation is streamlined via a single line added to ~/.config/opencode/opencode.json, referencing the GitHub repository. Alternatively, a local git clone can be used for development or offline use. Prerequisites include OpenCode installed and on your PATH, git, and Bun (which is bundled with OpenCode). The setup involves configuring opencode.json, potentially cloning the repository, and symlinking command files, followed by restarting OpenCode. Official OpenCode documentation is available at https://opencode.ai.

Highlighted Details

• Eleven Ported Skills: Includes code-review (multi-agent PR review), security-review (OWASP-bucketed audit with PoC requirement), feature-dev (seven-phase workflow), frontend-design (production-grade UI generation), and others like code-explorer, code-architect, skill-creator.
• Enhanced Review Workflows: code-review and security-review feature confidence-filtered cross-checks, reproduction scenarios, and mandatory category coverage or concrete PoC requirements.
• Direct Slash Command Invocation: Skills are available as slash commands (e.g., /code-review) that directly execute the inlined SKILL.md content, providing a robust user experience.
• Project Memory Management: Skills like agents-md-improver and agents-md-revise help audit and capture session learnings into project documentation (AGENTS.md/CLAUDE.md).

Maintenance & Community

The project is primarily maintained by waybarrios, with significant contributions and original work from Anthropic and Jesse Vincent (obra). Community links (Discord/Slack) are not specified in the README. Contributions for new skill ports and improvements are welcomed.

Licensing & Compatibility

The wrapper code and original contributions are MIT-licensed. Each ported skill cites its specific upstream source, with Anthropic projects also noted as MIT-licensed. This ensures broad compatibility for commercial use under standard MIT terms.

Limitations & Caveats

The effectiveness of multi-step skills can be dependent on the underlying language model's instruction-following capabilities; smaller models may provide less detailed responses. The project focuses on porting Claude Code's methodology and commands/agents, not its specific hooks or output styles. It is designed exclusively for the OpenCode environment.