Most "secret" Claude prompts are tone changes being sold as reasoning unlocks. I tested 40 of the most-shared codes in blind A/B trials over three months. 7 measurably change what Claude decides. 33 only change how it sounds. The cheat sheet has all 120 with before/after data so you stop wasting tokens on the placebos.
This is what every one of the 120 codes looks like inside. Real prompt, real before/after from Sonnet 4.6 and Opus 4.7, real token deltas, real failure modes. Picture yourself with this open on a second monitor next time you sit down to a hard prompt.
Forces decisive, committed answers. Eliminates "it depends" spirals when you need a recommendation.
Removes telltale LLM phrasing — "dive into", "landscape", "crucial". Output reads like a human wrote it.
Model critiques its own draft against a checklist before you see it. Catches ~60% of logic errors in one pass.
Pick the tier that matches how you'll use it. No subscriptions.
“Genuinely the first prompt resource I've used where the author admits half the famous tricks are placebo. The before/after examples settle arguments with my team in seconds.”
“Second real testimonial — reserved.”
40 pages. No email gate after download. No upsell.
