Rex Skills 🚀

English | 简体中文

📖 Overview

Welcome to Rex Skills, a curated collection of high-performance agent skills designed to supercharge your AI CLI experience.

"Open source is not just code; it's a culture of collaboration and innovation."

This repository hosts specialized skills (sub-agents) that extend the capabilities of your AI assistant. Whether you need rigorous prompt engineering, automated code audits, or hybrid model orchestration, these skills are built to production-grade standards.

📦 Available Skills

1. Anthropic 1P Prompt Optimizer

• Goal: Transform vague instructions into robust, production-ready prompts.
• Methodology: Follows Anthropic's "1P" interactive workflow (Role Prompting, XML separation, Chain of Thought, Guardrails).
• Use Case: When you need to rewrite system prompts, enforce JSON/XML outputs, or reduce hallucinations.

2. Code Review

• Goal: Automated, high-signal code review for Pull Requests and local diffs.
• Architecture: Multi-agent system (Guideline Checkers, Bug Detector, Context Analyzer) with confidence scoring.
• Features: Filters false positives (requires >80/100 confidence), integrates with GitHub/GitLab, and respects project-specific CLAUDE.md or .cursorrules.

3. Hybrid Executor

• Goal: The "right tool for the job" orchestrator.
• Function: Seamlessly delegates tasks across different AI CLIs:
  • Gemini CLI: For massive context analysis (logs, documentation).
  • Claude Code: For SOTA-level complex refactoring.
  • Codex/Main Agent: For task flow control.

4. OpenSpec Parallel Agents

• Goal: Orchestrate concurrent sub-agents for OpenSpec workflows.
• Features: Supports OPSX, legacy openspec, and Codex CLI commands. Safely handles multiple changes through dependency analysis.
• Use Case: When you need parallel execution of OpenSpec proposals, applications, or archival tasks while avoiding write conflicts.

5. Spec Kit Parallel Orchestrator

• Goal: Parallelize Spec Kit workflows (define, plan, implement).
• Methodology: Splits tasks into 3-6 parallel sub-agents with staged rollups.
• Use Case: When using /speckit.* commands or requesting Spec-driven concurrent development flows.

6. Superpowers Parallel Agents

• Goal: Accelerate "superpowers" tasks by splitting them into independent domains.
• Workflow: Schedules parallel agents for implementation and validation, ensuring non-overlapping edit scopes.
• Use Case: When tasks can be divided into different problem domains that can be solved and validated concurrently.

7. Review Loop Skill

• Goal: Automated two-phase code review loop for universal AI coding agents.
• Architecture: Implements task → independent CLI agent review → feedback addressing workflow.
• Features: Works with any Agent Skills compatible tool (Claude Code, Cursor, Codex, Gemini CLI, etc.), supports multiple reviewers (Codex, Claude, Gemini, OpenCode, Aider).

8. CCG Workflow OpenSkill

• Goal: Normalize ccg-workflow into an OpenSkill-compliant, gated workflow.
• Methodology: Command-to-lane routing (/ccg:*), phase gates, and Codex/Gemini parallel orchestration.
• Use Case: When migrating, executing, or optimizing ccg-workflow with OpenSkill protocol constraints.

9. Secrets Redaction

• Goal: Safely display .env/config/log output by masking secrets (API keys, tokens, passwords) before sharing.
• Use Case: When users ask to “show .env” or you need to debug config without leaking credentials.

🚀 Installation & Usage

To use these skills, you typically register the SKILL.md file with your AI CLI agent.

For example, to register the Code Review skill:

1. Navigate to the skill directory.
2. Point your agent to the SKILL.md file or copy its content into your agent's configuration.

# Example directory structure
rex-skills/
├── anthropic-1p-prompt-optimizer/  # Prompt Engineering
├── code-review/                    # Automated Auditing
└── hybrid-executor/                # Model Orchestration

🤝 Contributing

We love pull requests! If you have a new skill idea or an improvement to an existing one:

1. Fork the repo.
2. Create a new branch (git checkout -b feature/amazing-skill).
3. Commit your changes.
4. Open a Pull Request.

Please ensure your skill follows the SKILL.md metadata standard and includes adequate documentation.

---

📖 简介 (Overview)

欢迎来到 Rex Skills，这是一个精选的高性能 Agent 技能集合，旨在增强你的 AI CLI 体验。

“开源不仅仅是代码，更是一种协作和创新的文化。”

本仓库托管了多种专用技能（子代理），用于扩展 AI 助手的核心能力。无论你需要严谨的提示词工程（Prompt Engineering）、自动化的代码审计，还是混合模型的编排调度，这些技能都已达到生产级标准。

📦 可用技能 (Available Skills)

1. Anthropic 1P 提示词优化器

• 目标: 将模糊的指令转化为健壮的、生产就绪的提示词（Prompt）。
• 方法论: 遵循 Anthropic "1P" 交互式工作流（角色设定、XML 分隔、思维链 CoT、护栏 Guardrails）。
• 适用场景: 当你需要重写 System Prompt、强制要求 JSON/XML 输出格式，或减少模型幻觉时。

2. 代码审查 (Code Review)

• 目标: 针对 Pull Request 和本地 Diff 进行自动化的、高信噪比代码审查。
• 架构: 多代理系统（规范检查器、Bug 检测器、上下文分析器），基于置信度评分。
• 特性: 过滤误报（仅保留 >80/100 置信度的问题），集成 GitHub/GitLab，并遵守项目特定的 CLAUDE.md 或 .cursorrules 规范。

3. 混合执行器 (Hybrid Executor)

• 目标: “工欲善其事，必先利其器” —— 智能编排器。
• 功能: 在不同的 AI CLI 之间无缝委派任务：
  • Gemini CLI: 用于超长上下文分析（大规模日志、文档）。
  • Claude Code: 用于 SOTA 级别的复杂代码重构。
  • Codex/Main Agent: 用于整体任务流控制。

4. OpenSpec 并行代理 (OpenSpec Parallel Agents)

• 目标: 编排 OpenSpec 工作流的并发子代理。
• 特性: 支持 OPSX、旧版 openspec 和 Codex CLI 命令。通过依赖分析安全处理多个变更。
• 适用场景: 需要并行执行 OpenSpec 提案、应用或归档任务，且需避免写入冲突时。

5. Spec Kit 并行编排器 (Spec Kit Parallel Orchestrator)

• 目标: 并行化 Spec Kit 工作流（定义、计划、实现）。
• 方法论: 将任务拆分为 3-6 个并行子代理，并进行阶段性汇总。
• 适用场景: 使用 /speckit.* 命令或请求基于 Spec 驱动的并发开发流程时。

6. Superpowers 并行代理 (Superpowers Parallel Agents)

• 目标: 通过将任务拆分为独立领域来加速 "superpowers" 任务。
• 工作流: 调度并行代理进行实现和验证，确保编辑范围不重叠。
• 适用场景: 当任务可以划分为不同的问题域，且可以并发解决和验证时。

7. Review Loop 技能 (Review Loop Skill)

• 目标: 为通用 AI 编程代理提供自动化的两阶段代码审查循环。
• 架构: 实现任务 → 独立 CLI 代理审查 → 反馈处理工作流。
• 特性: 兼容任何 Agent Skills 工具（Claude Code、Cursor、Codex、Gemini CLI 等），支持多种审查者（Codex、Claude、Gemini、OpenCode、Aider）。

8. CCG Workflow OpenSkill

• 目标: 将 ccg-workflow 规范化为符合 OpenSkill 协议的可复用工作流。
• 方法论: 统一 /ccg:* 命令路由、阶段门禁、双模型并行编排与 OpenSpec 兼容约束。
• 适用场景: 需要迁移、执行或优化 ccg-workflow，并要求严格遵循 OpenSkill 规范协议时。

9. 敏感信息脱敏 (Secrets Redaction)

• 目标: 在分享 .env/配置/日志输出前自动脱敏（API Key、Token、密码等），避免泄露。
• 适用场景: 用户要求“查看 .env 内容”或需要排查配置问题但必须避免暴露凭证时。

🚀 安装与使用 (Installation & Usage)

要使用这些技能，通常需要将 SKILL.md 文件注册到你的 AI CLI Agent 中。

例如，注册 Code Review 技能：

1. 进入对应的技能目录。
2. 将 Agent 指向 SKILL.md 文件，或将其内容复制到 Agent 的配置中。

# 目录结构示例
rex-skills/
├── anthropic-1p-prompt-optimizer/  # 提示词工程
├── code-review/                    # 自动化审计
└── hybrid-executor/                # 模型编排

🤝 贡献 (Contributing)

我们非常欢迎 Pull Request！如果你有新的技能想法或想要改进现有技能：

1. Fork 本仓库。
2. 创建一个新分支 (git checkout -b feature/amazing-skill)。
3. 提交你的更改。
4. 发起 Pull Request。

请确保你的技能符合 SKILL.md 元数据标准，并包含充分的文档。

---

Built with ❤️ for the Open Source Community.