The creator of Claude Code's Claude setup
TL;DR Highlight
The developer who built Claude Code shares their own personal Claude setup — a rare look at the actual workflow behind the tool.
Who Should Read
Claude Code users and developers who want to optimize their AI coding workflow by seeing how an expert practitioner uses the tool.
Core Mechanics
- The post reveals the Claude Code author's actual CLAUDE.md configuration, custom commands, and workflow patterns — not theoretical best practices but real day-to-day usage.
- Key insight: the author uses very minimal global instructions and relies heavily on per-project CLAUDE.md files, letting context be specific rather than generic.
- Tool permissions are configured conservatively even by the author — default to asking rather than auto-executing for filesystem-modifying operations.
- The author uses Claude Code primarily for exploration and refactoring of existing codebases rather than greenfield generation.
- Custom slash commands are used for common multi-step workflows (e.g., creating PRs, running test suites) — reducing repetition without over-automating.
- The author is candid about what Claude Code is bad at and where they still prefer manual coding — realistic expectations matter more than hype.
Evidence
- The post was highly upvoted on HN with commenters praising its candor — a refreshing contrast to vendor marketing about AI tools.
- Multiple users noted that seeing the actual config files was more useful than any tutorial, because it shows the real tradeoffs made in practice.
- Debate about whether the author's conservative permission setup is appropriate for other users — some argued more automation is worth the occasional mistake.
- Several Claude Code users reported updating their configurations based on the post, particularly around CLAUDE.md structure.
How to Apply
- Review your own CLAUDE.md against the author's example — are you over-specifying global rules that should be project-local?
- Audit your tool permissions: are there operations where you've enabled auto-execution that you'd actually prefer to review before running?
- Consider moving common multi-step workflows to custom slash commands rather than repeating them in every session.
- Use this as a calibration check: if the person who built the tool doesn't use certain features heavily, that's a signal about which features are actually reliable.
Terminology
Related Papers
Show HN: OpenKnowledge – open source AI-first alternative to Obsidian/Notion
Git 기반 동기화와 Claude/Codex/Cursor 연동을 내장한 로컬 우선 마크다운 에디터로, AI 에이전트의 두 번째 뇌(LLM Wiki)로 활용할 수 있는 오픈소스 도구다.
The Unfireable Safety Kernel: Execution-Time AI Alignment for AI Agents and Other Escapable AI Systems
AI 에이전트가 자신의 안전장치를 우회할 수 없도록, 에이전트 프로세스 바깥에 수학적으로 증명된 강제 통제 게이트를 배치하는 아키텍처
RubyLLM: A Ruby framework for all major AI providers
OpenAI, Claude, Gemini 등 주요 AI 프로바이더를 단일 인터페이스로 통합한 Ruby 프레임워크로, Rails 통합과 에이전트 기능까지 지원해 Ruby 개발자가 AI 기능을 빠르게 붙일 수 있다.
Qwen-AgentWorld: Language World Models for General Agents
Alibaba Qwen 팀이 AI 에이전트가 행동 결과를 미리 시뮬레이션할 수 있는 'Language World Model'을 공개했다. 에이전트 훈련과 실행 경로 검증에 새로운 패러다임을 제시하는 연구다.
SHERLOC: Structured Diagnostic Localization for Code Repair Agents
버그 위치만 알려주는 게 아니라 '왜, 어떻게 고쳐야 하는지'까지 진단 리포트를 생성해서 코드 수정 에이전트의 성능을 높이는 training-free 프레임워크
Show HN: peerd – AI agent harness that runs entirely in your browser
백엔드 서버 없이 Chrome/Firefox 확장 프로그램으로만 동작하는 AI 에이전트 실행 환경으로, 브라우저 탭을 직접 조작하고 WASM Linux VM까지 구동할 수 있어 프라이버시와 보안을 동시에 챙길 수 있다.