** 과학 및 공학을위한 휴대용 AI 시약 워크 **

Genesis Skills는 과학적 발견과 엔지니어링 워크플로우를 위한 재사용 가능한 Agent Skills 컬렉션입니다. Genesis Mission의 일부, 카탈로그는 Claude Code, Codex, Cursor, OpenCode, Gemini CLI 및 LangChain Deep Agents와 함께 사용할 수 있습니다. 개인 기술은 여전히 클라이언트 별 기능 또는 외부 도구에 의존 할 수 있습니다.

Skills encode 도메인 전문성, 과학 데이터 소스에 연결하고, 계산 연구에 대한 재현성 패턴을 제공합니다.

---

기술이란?

Skills are markdown files (SKILL.md]) with YAML frontmatter that teaching a AI Agent 도메인 별 작업을 처리하는 방법. 각 기술 제공:

• Domain 전문성 - 문학, HPC, 플라즈마 시뮬레이션, 과학 데이터 메타데이터 및 AI-시스템 평가에 대한 전문 지식 -Tool Access - 특정 도구 사용 권한(Bash 명령, 파일 운영, 웹 API)
• **Structured 워크플로우 ** - 복잡한 과학 작업을 위한 단계별 지침
• ** 지원 리소스 ** - 스크립트, 템플릿 및 참조 자료 에이전트는 사용할 수 있습니다

에이전트는 자연 언어 요청에서 기술을 선택할 수 있습니다.

Genesis Skills

Portable AI-agent workflows for science and engineering

Genesis Skills is a collection of reusable Agent Skills for scientific discovery and engineering workflows. Part of the Genesis Mission, the catalog can be used with Claude Code, Codex, Cursor, OpenCode, Gemini CLI, and LangChain Deep Agents. Individual skills may still depend on client-specific features or external tools.

Skills encode domain expertise, connect to scientific data sources, and provide reproducible patterns for computational research.

---

What Are Skills?

Skills are markdown files (SKILL.md) with YAML frontmatter that teach an AI agent how to handle domain-specific tasks. Each skill provides:

• Domain expertise — Specialized knowledge for literature, HPC, plasma simulation, scientific-data metadata, and AI-system evaluation
• Tool access — Permissions to use specific tools (Bash commands, file operations, web APIs)
• Structured workflows — Step-by-step instructions for complex scientific tasks
• Supporting resources — Scripts, templates, and reference materials the agent can use

Agents can select a skill from a natural-language request. Client-specific menus and explicit invocation shortcuts vary by agent.

---

Quick Start

There’s nothing to compile: clone the repo, then expose the skills to your agent. The repository groups skills by domain (skills///), while agent clients generally expect each skill to be a direct child of a configured skills directory. unpack.sh flattens that layout. LangChain Deep Agents receive a catalog path in code instead of discovering a user-level directory automatically.

git clone https://gitlab.osti.gov/genesis/genesis-skills.git
cd genesis-skills

Mirror repository: The repository https://gitlab.osti.gov/genesis/genesis-skills.git is a read-only mirror of the upstream development repository https://github.com/AI-ModCon/genesis-skills. Please use GitHub for development, forks, issues, and pull requests.

Method 1 — Flatten skills into place (unpack.sh)

Choose the skills directory documented by your client, then let unpack.sh flatten every skill into it:

export SKILLS_DIR="$HOME/.agents/skills"  # Codex, Cursor, OpenCode, Gemini CLI
# export SKILLS_DIR="$HOME/.claude/skills"  # Claude Code
./unpack.sh --target "$SKILLS_DIR" --mode copy

Use --mode symlink for live links back to the clone, ./unpack.sh hpc-skills for one domain, ./unpack.sh --list to preview, or ./unpack.sh --help for all options. The legacy default remains .claude/skills; an explicit --target is the client-neutral choice. Existing same-named destination entries are replaced.

Method 2 — Point the skill-search skill at the catalog

Expose just the bundled skill-search skill; it discovers the whole nested catalog