Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Claude Code Starter Kit

A drop-in CLAUDE.md behavioral contract + a verify-before-done guardrail skill for Claude Code. Free, MIT-licensed, no signup to use.

Your Claude Code setup is probably fighting you — forgetting context, editing the wrong files, calling broken work "done." This is the single highest-leverage fix: copy two files in, tune, done.

Why this exists

A great model with a sloppy harness is a junior dev with amnesia. The two pieces in this repo are the highest-leverage parts of a production setup:

  • CLAUDE.md — a behavioral contract (not a wish list): response style, opinionated-peer behavior, decision-fork handling, a "done means done" rule, and root-cause error prevention. The agent re-reads it every session.
  • skills/verify-before-done.md — a guardrail skill: the agent must run a real check and quote the output before claiming "done", "fixed", or "passing." Evidence before assertions.

What's inside

File What it does
CLAUDE.md Drop-in behavioral contract — place, then tune.
skills/verify-before-done.md Example guardrail skill: verify before claiming success.
START-HERE.md 3-step, ~10-minute quick start.

Quick start (~10 min)

  1. Place the contract. Copy CLAUDE.md to your project root (project rules) or ~/.claude/CLAUDE.md (global). Start a fresh Claude Code session so it loads.
  2. Add the skill. Copy skills/verify-before-done.md to ~/.claude/skills/verify-before-done/SKILL.md.
  3. Tune it. Edit the contract to match how you actually work — response style, your stack, your conventions.

Want the full setup?

This is the free starter. The full Claude Code Setup Playbook adds the parts that take a setup from "works" to "compounds":

  • A skill library that scales (structure, naming, routing — not 300 dead files)
  • Model routing that sends cheap work to cheap models automatically
  • Persistent memory + a knowledge graph so the agent never re-researches your own work
  • Guardrail hooks (verify-before-done, pre-build recall) wired and enforced
  • A publication-quality PDF/HTML output pipeline

Grab the fuller free kit + get on the list for the full Playbook: https://expressive446.gumroad.com/l/qlypgs

Built and run by someone who ships with this stack daily.

License

MIT — free to use, adapt, and share. See LICENSE.


Claude is a trademark of Anthropic PBC. This project is not affiliated with, endorsed by, or sponsored by Anthropic.

About

Free drop-in CLAUDE.md behavioral contract + a verify-before-done guardrail skill for Claude Code.

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors