Act on this, mostly signal

Free ruleset for cutting Claude usage on a one-person business

What to do

Adopt the paste-in rules, with CONTEXT.md and one task per chat, and measure your own before and after usage.

Why

The claimed 70% token saving is modeled, not measured, so verify it yourself.

What the research found

Free open-source paste-in ruleset (CONTEXT.md, one task per chat, model check, handoffs) for cutting Claude usage; the claimed 70% token saving is modeled arithmetic, not measured, and the author asks for real before/after numbers.

Claims and how well the evidence supports them

The README models about 70% fewer tokens for a typical day of work 95%
The fix is a CONTEXT.md file that every chat reads first 93%
The fix is free and open source on GitHub 85%
The paste-in prompt is about 400 words and needs no code 97%

Evidence

reddit.com https://reddit.com/r/ClaudeAI/comments/1xat7zo/i_kept_hitting_my_claude_limit_running_a
github.com https://github.com/louddesigns99-ui/claude-limit-saver

Actionability rated low.