Universal Domain Rules Discover Skill — Testing skill for Claude Code

Universal Domain Rules Discovery

A vendor-neutral skill for AI agents to discover, extract, compare, validate, and document application-specific domain rules before QA execution.

This repository is designed to work with **any AI agent capable of reading instructions and accessing project artifacts**. It does not depend on a proprietary skill runtime. The canonical behavior lives in `SKILL.md`; platform-specific adapters only explain how to load it.

Why universal instead of one native format?

There is no single skill format supported by every agent. This repository solves that by separating:

1. Canonical protocol — SKILL.md
2. Machine-readable contract — manifest.yaml and JSON Schemas
3. Platform adapters — files under adapters/
4. Portable output artifacts — Markdown and JSON templates
5. Validation tooling — scripts/validate.py

Supported usage models

• OpenAI Codex and agent runtimes
• Claude Code
• Cursor
• Gemini CLI
• GitHub Copilot coding agents
• IDE agents
• CI/CD agents
• Autonomous repository agents
• Custom LLM agents
• Any system that accepts a system prompt, repository instruction file, or task policy

Core promise

The skill guarantees a repeatable **process**, not identical wording. Every compliant agent must:

• trace every rule to evidence;
• separate confirmed rules from hypotheses;
• expose conflicts instead of silently resolving them;
• record gaps and uncertainty;
• generate domain-specific QA scenarios;
• avoid claiming complete coverage without evidence.

Quick start

Generic agent

Give the agent this instruction:

Read SKILL.md as a mandatory execution protocol. Analyze the target repository and available product evidence. Produce every required artifact under domain-discovery/. Do not modify application code. Validate JSON outputs against the schemas in schemas/ before finishing.

Repository-level installation

Copy the repository contents into one of these locations: