Skill · CLAUDE.md rule

A CLAUDE.md skeleton — turn a description into operating law

A minimal frame to paste into CLAUDE.md that tells the agent what to default to, when to stop and ask, and where the danger is.

TypeCLAUDE.md rule

LanguageĐọc bản tiếng Việt →

You write a CLAUDE.md, describe the project beautifully, and the next session you're still repeating every item: "remember this voice," "don't touch that folder," "check before you say done." It feels like the file does nothing. The problem isn't a lazy agent — it's that you wrote a description, and a description is something the agent reads for context and then freely does its own thing anyway.

A CLAUDE.md that earns its keep, whatever the project, commands: default to this, stop and ask there, this zone needs care, done means verified like so. The frame below is as lean as it gets while still covering those five. Drop the flattering prose, keep the parts that force behavior:

✕ Written like a README

Describes the project, architecture, stack

▼

Agent reads it for context, then does its own thing

▼

Every session you repeat yourself by hand

✓ Written like law

Required defaults · stop-and-ask triggers · danger zones

▼

Agent holds the line without being told again

▼

Set it once, use it every session

Paste this into CLAUDE.md and fill the brackets. The two highest-value sections are "Stop and ask" and "Before you say it's done" — that's where one line of law catches most of the wrong guesses and the unchecked "Done!" Keep it short; each line is an instruction, not a paragraph.

Paste into CLAUDE.md

# <Project> — CLAUDE.md

## What this is
<one line: what it does, who it's for — keep it to a line>

## Conventions (follow these by default, without being told)
- <naming / structure / style the agent must default to>
- <the thing you keep having to repeat — write it here once>

## Stop and ask — do not guess
- When a requirement is ambiguous, ask before choosing.
- Before any destructive or irreversible action.

## Danger zones (handle with extra care)
- <files/areas where a mistake is expensive>

## Before you say it's done
- <how to verify: run it / show the real output / the check that must pass>

Prefer your agent install it?

Add this rule to my CLAUDE.md, verbatim:

# <Project> — CLAUDE.md

## What this is
<one line: what it does, who it's for — keep it to a line>

## Conventions (follow these by default, without being told)
- <naming / structure / style the agent must default to>
- <the thing you keep having to repeat — write it here once>

## Stop and ask — do not guess
- When a requirement is ambiguous, ask before choosing.
- Before any destructive or irreversible action.

## Danger zones (handle with extra care)
- <files/areas where a mistake is expensive>

## Before you say it's done
- <how to verify: run it / show the real output / the check that must pass>

Did this help?

Tags#claude.md #setup #skill

Other skills

01Verify Before You Call It Done 02Plan First, Then Code 03Map the Blast Radius Before You Edit 04If It's Deterministic, Don't Make It Guess 05Diagnose From the Symptom, Not the Guess 06Make It Argue Against Its Own Plan 07Hand Off to Your Next Session 08A Review That Doesn't Rubber-Stamp 09Run to the gate, then stop — not step-by-step 11The weekly review prompt — five minutes that turn experience into skill 12Split the work into focused slices, each with clean context 13Is this even worth an agent — ask before you open the chat 14Probe with a small task before you trust it in a new domain

Back to skills

The author

Each story here wraps a lesson paid for in full.

craftagentsomeone building and learning at once

What are you building with agents? Want to trade notes, push back, or build something together — drop a line.

Write hello@craftagent.cloud

58pieces12clustersVI·ENbilingual

✕ Written like a README

✓ Written like law

Get new pieces by email