Claude Booster

**🇬🇧 English** | [🇷🇺 Русский](README.ru.md)

Claude Booster

**Stop re-teaching Claude Code the same things every morning.**

Claude Code out of the box has no memory across sessions, no institutional learning, no cross-project knowledge transfer. By week three of daily use, you notice:

• You re-explain the stack, the conventions, the failure modes — every session.
• Claude reimplements a helper you already have, because it didn't grep first.
• Same clarifying questions, day after day. ("npm or pnpm?" — you answered yesterday.)
• A hook silently stopped firing and you discovered it 3 weeks later.
• Every new project starts at zero. Hard-won lessons from the old one don't transfer.

Claude Booster turns those sessions into a compounding asset. One `python install.py` on any Mac or Linux box and your Claude Code starts **remembering, learning, and auditing itself**.

---

🚀 Start here (newcomer's guide)

New to Claude Booster? You don't need to read the whole changelog below. Here's the 5-minute version.

1. Install (one command)

git clone https://github.com/demetrius2017/claude-booster.git
cd claude-booster && python install.py

That's it. Booster installs a set of rules, hooks, and slash commands into `~/.claude/`. Your next Claude Code session already behaves differently — it reads its own memory at startup and enforces a quality pipeline on itself.

2. Your first session

Step	Type this	What happens
Open a session	/start	Claude reads yesterday's handover, the knowledge base, and health telemetry, then proposes a plan. No more re-explaining your stack every morning.
Do coding work	/go	The task runs through a gated, cross-provider pipeline (design → challenge → build → independent read-only evidence → diff review → final deploy regression gate). The verdict is an evidence receipt, not "looks good to me."
Close the session	/handover	A str