Claude Md Templates — Development skill for Claude Code

CLAUDE.md Templates

Ready-to-use templates and skills for writing effective CLAUDE.md files — the context files that make AI coding agents work like you.

**The golden rule** — Write only what the agent cannot discover on its own. Every other line is noise that costs tokens and hurts focus.

Why this matters

Agents follow your instructions literally. Bad instructions — generic advice, discoverable information, verbose descriptions — actively make things worse. Research confirms that verbose, auto-generated context files reduce performance while increasing cost. The fix is simple: write less, but write what matters.

The files you need

File	Location	Scope	Target
`global-CLAUDE.md`	~/.claude/CLAUDE.md	All projects	~30 lines
`project-CLAUDE.md`	./CLAUDE.md (repo root)	This project	~40 lines
`.claude/rules/*.md`	.claude/rules/	Targeted file patterns	~10 lines each

Global — your engineering identity

[`global-CLAUDE.md`](./global-CLAUDE.md)

• Philosophy — DRY, YAGNI, simplicity over cleverness
• Response language and style
• Code conventions (naming, strictness)
• Preferred tools (package manager, test runner)
• Git commit conventions
• Workflow rules

Project — this repo's context

[`project-CLAUDE.md`](./project-CLAUDE.md)

• Philosophy — this project's design principles
• What the project does (1-2 sentences)
• Non-obvious stack choices
• Commands with special setup
• Architecture decisions
• Patterns to follow
• Explicit "do not" guardrails (with reasons)

Why Philosophy matters

The Philosophy section shapes **how** the agent thinks, not just what it does. Without it, the agent defaults to generic patterns. With it, the generated code feels like **yours**.

Level	What it captures	Example
Global	Your engineering identity	"YAGNI: only implement what's needed NOW"