⚡ AI FORGE
Penloom Studio
← All products
An agent with no project context makes confident, wrong decisions — the wrong test command, the wrong framework convention, a "fix" that quietly reintroduces last week's bug. CLAUDE.md is the file it reads at the start of every session. Get it right once and you stop re-explaining your project every time you open a chat.

Sound familiar?

  • It keeps "forgetting" your conventions. The wrong test command, the wrong stack idiom, the bug you already fixed — reintroduced.
  • You have a CLAUDE.md — a 400-line wall of text. The model half-reads it and ignores the rules that matter.
  • Every teammate's agent behaves differently. No shared file, no shared behavior.
  • You retype the same three corrections every session. Because the one file that would fix it isn't set up.

What's inside — 12 files

Written from hundreds of real agent cycles, not theory. Copy a template, delete what you don't need, ship.

📄

6 copy-paste templates

A 20-line minimal starter, a fully annotated full template, and pre-filled versions for Node/TypeScript, Python, and React/Vite — plus an AGENTS.md for multi-tool teams (Cursor, Codex, Copilot, Aider, Windsurf).

⚡

QUICKSTART — set up in minutes

Copy the minimal template into your repo root, rename it CLAUDE.md, fill in a handful of blanks. Your next session starts already knowing how your project works.

🖨️

The 1-page cheatsheet

The rules that actually change agent behavior versus the ones it quietly ignores — short, specific, and structured the way agents follow. Print it and keep it next to you.

🔍

2 annotated walkthroughs

A side-by-side good-vs-bad (rules agents follow vs. ignore) and a real CLAUDE.md walked

How it works — 30 seconds to a working config

Copy. Drop CLAUDE.minimal.md into your repo root and rename it CLAUDE.md.
Fill in the blanks. Build and test commands, your conventions, and the do-nots — the cheatsheet tells you exactly which lines earn their keep.
Open a session. The agent now knows your project from line one — no re-explaining.
Grow it. Swap in the full or stack-specific template as your rules pile up, and share AGENTS.md so every tool on the team behaves the same.

Stop re-explaining your project every session.

your next agent session already knows the rules.

Coming soon: pay-per-call access pending approval

Questions, answered

What exactly do I get?

Twelve files: six CLAUDE.md templates (a 20-line minimal starter, a full annotated template, and pre-filled Node/TypeScript, Python, and React versions, plus a cross-tool AGENTS.md), a QUICKSTART, a one-page CHEATSHEET, two annotated examples (good-vs-bad and a real-repo walkthrough), and a README plus a dated CHANGELOG. Delivered to your email automatically after checkout.

Does it only work with Claude Code?

No. CLAUDE.md is the file Claude Code reads first, but the included AGENTS.md is the shared version for multi-tool teams — Cursor, Codex, Copilot, Aider, and Windsurf all read it, and one settings line points Gemini CLI at it too.

Do I need to know how to write one already?

No. Copy the 20-line minimal template into your repo, fill in a few blanks from the cheatsheet, and you're done. The full and stack-specific templates are there for when your rules grow.

Is it a subscription?

What if it isn't what I expected?

Email within 14 days. If it doesn't deliver or isn't as described, we'll make it right with a replacement or refund.

The CLAUDE.md Starter Kit is a set of copy-paste templates and reference files, current as of July 2026 and updated for free as tools change. It is an independent product and is not affiliated with or endorsed by Anthropic. No config guarantees a perfect agent — the point is a short, specific file your tools actually follow.

Coming soon: pay-per-call access pending approval