재사용 가능한 에이전트 워크플로우 및 프롬프트의 공개 컬렉션.

기술을 설치하려면, 당신이 원하는 기술에 에이전트를 포인트하고 그것을 설치하도록 요청.

기술

• /agent-browser — automate browser, 웹사이트, Electron 앱 워크플로우는 Agent-browser CLI.
• /autoreview - 커밋 또는 배전하기 전에 구조화 된 코덱 또는 클로드 코드 리뷰 실행.
• /brainstorm - 방향을 선택하기 전에 아이디어를 생성하고 탐구합니다.
• /codebase-quiz - 실제 구현 작업과 사전 코딩 세션을 대화형 퀴즈로 HTML 수업으로 전환합니다.
• /create-cli - 디자인 명령줄 인터페이스, 플래그, 도움말, 출력, 오류, 설정 및 건조 실행 동작.
• /frontend-design - 일반적인 레이아웃 대신 독특한 광택 frontend UI를 구축.
• /grill-me - 계획, 디자인, 접근을 스트레스 테스트에 대한 사용자 relentlessly 인터뷰.

각각의 삶은 skills/ /SKILL.md.

Skills는 재사용 가능한 워크플로우입니다. Matt Pocock의 컨벤션은 다음과 같습니다. user-invoked Skill set disable-model-invocation: true]; model-invoked Skill omit that field and keep the model-facing description.

# Prompts의

• prompts/security-audit.md - 단계별, 보고 전용 보안 감사 프롬프트.

Agent Scripts

A public collection of reusable agent workflows and prompts.

To install a skill, point your agent to the skill you want and ask it to install it.

Skills

• /agent-browser — automate browser, website, and Electron app workflows with the agent-browser CLI.
• /autoreview — run structured Codex or Claude code reviews before commit or ship.
• /brainstorm — generate and explore ideas before choosing a direction.
• /codebase-quiz — turn real implementation work and prior coding sessions into a visual HTML lesson with an interactive quiz.
• /create-cli — design command-line interfaces, flags, help, output, errors, config, and dry-run behavior.
• /frontend-design — build distinctive, polished frontend UI instead of generic layouts.
• /grill-me — interview the user relentlessly to stress-test a plan, design, or approach.

Each one lives in skills//SKILL.md.

Skills are reusable workflows. Matt Pocock’s convention is: user-invoked skills set disable-model-invocation: true; model-invoked skills omit that field and keep a model-facing description.

Prompts

• prompts/security-audit.md — phased, report-only security audit prompt.

Prompts are manual, explicit-use instructions. They live outside skills/ so agents do not invoke every workflow by themselves.

Codex CLI does not currently use disable-model-invocation to control skill invocation. For Codex, implicit skill use is controlled by agents/openai.yaml with policy.allow_implicit_invocation; keep Codex-only manual workflows in prompts/.

Acknowledgements

This collection builds on ideas from Matt Pocock and Brian Madison’s agent workflow methodology. I love their work.

• /agent-browser is from Vercel Labs’ agent-browser.
• /autoreview and /create-cli are from Peter Steinberger’s agent-scripts.
• /frontend-design is from Anthropic’s skills.
• /grill-me is a standalone adaptation of Matt Pocock’s grill-me and grilling skills.
• /brainstorm is an original standalone workflow inspired by BMAD’s brainstorming approach.