WorkspaceBrowse, save, and manage patterns without leaving the shell

The Solo CLAUDE.md

A single CLAUDE.md file under 200 lines with project context, commands, and conventions. The Anthropic-recommended starting point.

minimalWeb AppFeatured
Q
Quzr27@quzr27

software engineer

1 followerquzr27
AI Tools
claude-code
File structure2 files
13
Pros
Zero setup — run /init and start shipping in minutes
Maximizes instruction-following quality by keeping token count low
Easy to maintain — one file, one owner, no coordination overhead
Cons
Does not scale past ~200 lines without degrading instruction adherence
No tool-specific scoping — all instructions load on every session
Avoid if
Your team has more than 2 people working on the same repo
Your CLAUDE.md already exceeds 200 lines and keeps growing
You use multiple AI tools that need different configurations

Description

The simplest effective setup for Claude Code. Anthropic's own docs recommend keeping CLAUDE.md concise and human-readable — ideally under 200 lines. HumanLayer's production CLAUDE.md is just 60 lines. The key insight from research: frontier LLMs can reliably follow ~150–200 instructions total, and Claude Code's system prompt already uses ~50 of those. Every line you add to CLAUDE.md competes for attention. Start with what Claude gets wrong, not a comprehensive manual. Use /init to generate a starter, then prune aggressively. If Claude keeps ignoring a rule, your file is probably too long.

Who is this for

Solo developers or small teams starting with Claude Code. Best for projects where one person holds the full context.

Why files are split this way

Everything fits in a single file because solo developers can hold full project context. Splitting into multiple files creates maintenance overhead without benefit at this scale. If the file exceeds 200 lines, that is the signal to graduate to a more structured pattern.

What changed from the previous setup

No AI configuration at all — Claude Code was used with its defaults and no project-specific context.