This is a reference document of 120 tested prompt codes — short prefixes like L99 or /ghost that you type at the start of a Claude prompt. Nothing installs. Looking for the free Skills library? →
You’ve been using Claude Code on hard mode.
100 tested prompt patterns that change how Claude responds.
Not hidden API commands — community-discovered shortcuts that reliably trigger specific behaviors. Most developers don’t know they exist. The ones who do ship in half the time.
Read 3 full codes before you pay a cent.
These are unedited excerpts from the paid cheat sheet. Same format as the other 97 codes inside. Try them in Claude and decide for yourself.
L99Pushes Claude to maximum reasoning depth. Less hedging, more opinion, longer answers.
Should I rewrite my Express backend in Go?
L99 Should I rewrite my Express backend in Go?
Claude stops giving you 'it depends' and starts giving you a real answer. Default mode hedges because it's trying not to be wrong; L99 explicitly tells it to commit. You'll get a 4-5 paragraph response with a clear recommendation, the specific tradeoffs, and the conditions under which it would change its mind. The hedging instinct doesn't fully die but it gets quieter.
Don't L99 questions where the right answer genuinely is 'it depends' — like 'what's the best database?'. L99 will manufacture a confident answer that sounds smart but is wrong half the time.
/deepthink/blindspots/blindspotsAsks Claude to find what you're missing or assuming, not what you're asking.
Review my plan to migrate from Postgres to MongoDB.
/blindspots Review my plan to migrate from Postgres to MongoDB.
Most reviews check 'is the plan internally consistent?'. /blindspots asks 'what is this plan assuming that might not be true?'. You'll get things like: 'You're assuming your data is mostly read-heavy, but your schema suggests heavy writes', or 'You haven't mentioned how you'll handle existing foreign-key constraints'. It surfaces the questions you didn't think to ask.
Avoid on small decisions. Blindspot analysis is 5x the work for an answer to a question you weren't asking — pointless for 'should I name this function camelCase or snake_case'.
INVERTWORSTCASE/devil/ghostStrips every AI tell — em-dashes, hedging, 'I hope this helps', the works.
Rewrite this so it sounds more natural: [paragraph]
/ghost [paragraph]
Claude removes the three things every AI detector flags: balanced sentence rhythm, transitional phrases like 'furthermore' and 'however', and the tic of opening with 'In conclusion'. You get prose that reads like a tired human typed it at 11pm.
Don't use it for technical docs or anything that needs to sound authoritative. /ghost makes everything feel a little casual and a little broken on purpose. Wrong tone for a security advisory.
/voice/punchThose are 3 of 100.
The other 97 get the same treatment. Plus 10 workflow playbooks and 20 bonus power prompts in Pro.
See pricing ↓Why not just Google these codes?
You can. Here’s the math on what that actually costs you.
- ✗20+ hours googling 'claude prompt codes' across Reddit, Twitter, dead Medium posts
- ✗~40 codes scraped, half duplicates, half stale (posted before Claude 3.5)
- ✗No examples — you know the name, not what it changes in the response
- ✗No failure modes — you learn them by wasting prompts in production
- ✗No combos — you never find out /punch + /trim + /raw is the Slack combo
- ✗No workflows — you have a list, not a playbook
- ✗No updates — when new codes appear, you hunt them down again
Total cost: 20-30 hours plus every wrong answer Claude gave you along the way.
- ✓120 codes personally tested over 3 months — the ones that didn’t work aren’t in here
- ✓Before/after examples for every code — what actually changes, not just what it does
- ✓When-NOT-to-use warnings — the failure modes nobody else documents
- ✓Code combos that stack — which codes amplify each other and which conflict
- ✓10 real-world workflow playbooks — recipes, not code lists
- ✓20 reusable bonus prompts (Pro) — full copy-paste templates
- ✓Lifetime updates — new codes added automatically. Bookmark once.
Total cost: one-time payment. Yours forever. Lifetime updates.
Pick your tier
One-time payment. Lifetime updates. Pay via PayPal — any card, anywhere.
Lite
Try it without committing.
- 50 most-used prompt codes
- Before/after examples
- Markdown file download (.md)
- Live web access · lifetime updates
- When-NOT-to-use warnings
- Code combos that stack
- 10 real-world workflow playbooks
- 20 bonus power prompts
- 30-day email channel with Samarth
Full
All 120 codes with deep analysis. Most buyers choose this.
- All 100+ prompt codes
- Before/after examples
- When-NOT-to-use warnings
- Code combos that stack
- 10 real-world workflow playbooks
- Markdown file download (.md)
- Live web access · lifetime updates
- 20 bonus power prompts
- 30-day email channel with Samarth
Pro
Everything in Full + Deep Dive + 30-day direct line to the builder.
- Everything in Full
- Deep Dive: verified-test analysis of the highest-value codes
- Reasoning-shifter classification (which codes actually work)
- Lookalike-placebo warnings (the codes that look similar but don't)
- 20 bonus power prompts (reusable templates)
- 30-day direct email channel with Samarth
- Early access to new codes before public release
- Lifetime priority on new bonus content
Secure checkout · Instant delivery via email · No subscription · No upsells
My instinct is that if you under price something, people don’t value it as much. I have already been testing and have been really impressed. You are literally giving this away and the value is enormous.
Here’s exactly what you’re getting
No fluff. Just the actual contents.
100 tested prompt codes, organized into 10 categories
Writing & Style, Thinking & Reasoning, Coding & Technical, Power Commands, Hidden Modes, Output Formatting, Research & Analysis, Productivity Hacks, Communication, and Bonus / Experimental. Every code has been tested. The ones that didn't work consistently aren't in here.
Before/after examples for every single code
Not 'here's what /ghost does'. Actual side-by-side: the prompt you'd normally type, the prompt with the code, and what concretely changes in Claude's response.
When NOT to use each code (the failure modes)
Every code has situations where it makes things worse. /punch ruins legal text. /godmode wastes context on simple questions. PERSONA fails without specifics. The cheat sheet tells you when to skip each one.
The code combos that stack
Most codes are 2x more powerful when combined. /punch + /trim + /raw turns rambling Slack messages into 3-line wins. The cheat sheet maps which codes stack and which conflict.
10 real-world workflow playbooks
Full + ProNot generic advice — actual workflows with named steps. 'Turn an 11pm rant into a Slack message'. 'Debug a flaky test in under 30 minutes'. Each chains 3-5 codes together with the exact sequence.
20 bonus power prompts (Pro only)
Pro onlyDifferent from the codes. Full reusable prompt templates you copy, fill in the brackets, and run. Meeting notes → action items. Bug rant → Linear ticket. Cold email writer. Negotiation script. 20 of them.
30-day direct email line to Samarth (Pro only)
Pro onlyFor a month after purchase, email directly with questions, prompt requests, or anything else. Not a chatbot — a real human (me) who tested every code in here.
Markdown file, live web access, lifetime updates
Markdown attached to your purchase email (yours forever, even offline), live web link that stays current as new codes are added, and a download endpoint that works from any device. One email recovery if you lose the link.
Three reasons you’re hesitating
And why each one is wrong.
“I can find these codes for free if I just google.”
You can find some of them, scattered across reddit threads and tweets, with no examples and no idea which ones still work. I spent 3 months testing every code in this cheat sheet so you don’t have to. The codes that didn’t survive testing aren’t in here. The ones that did have examples and warnings. Your time is worth more than $10.
“I don’t use Claude that much.”
Then start using it more — but use it well. The cheat sheet isn’t reference material you read once. It’s a working document you come back to every week. Buyers who use 5 codes on day one have already gotten their money’s worth.
“What if I already know some of the codes?”
You probably know 10-15 of the 120. The value isn’t the list — it’s the before/after proof showing which ones actually shift Claude’s reasoning vs. which ones only change formatting, plus the failure modes and combos. You’ll learn more about the codes you already use than about the ones you don’t.
Two months from now,
you’ll wish you bought it today.
Every week you don’t have these codes is a week of asking Claude the wrong way.
Every prompt that returns a hedged, generic, useless answer is a prompt you could have run with L99, PERSONA, or /nofilter and gotten the answer the first time.
Stop using Claude on hard mode.
Get the cheat sheet — from $10 →Lifetime updates · Instant delivery · One-time payment
Questions
Can't I just Google these codes for free?
You can find the code names for free anywhere. What you can't find is which ones actually work (we tested all 120), which ones are placebo (about 70%), what the failure modes are, which combinations compound, and the real before/after output text. The cheat sheet is the test results, not the list. The list is free at clskillshub.com/prompts.
How is this different from asking ChatGPT?
Ask ChatGPT what the best Claude prompt codes are and you'll get a generic list that mixes working codes with placebo ones. ChatGPT hasn't tested them. We ran each code through a testing harness: same prompt, with and without the prefix, 3 runs each, across 5 task types. The cheat sheet is the data from that testing, not a list someone made up.
What's the difference between Full and Pro?
Full gets you all 120 codes with before/after proof, when-NOT-to-use warnings, code combos, and 10 real-world workflow playbooks. Pro adds the Deep Dive section with verified-test analysis of the highest-value codes, 20 reusable bonus prompt templates, and a 30-day direct email channel with the author.
What format is it delivered in?
Three ways at once: a private unlock link to a beautifully formatted web page, a markdown file attached directly to your purchase email, and a download link that works from any device. You always have a copy, even if you lose the email.
Do I get future updates?
Yes. New codes are tested and added regularly. Your unlock link works forever and always shows the latest version.
What if I lose my unlock link?
Visit clskillshub.com/cheat-sheet/recover and enter the email you used. We'll resend everything instantly.
Can I upgrade from Full to Pro later?
Yes. Email samarth@clskillshub.com with your purchase email and we'll send a one-time upgrade link for the price difference.
Are these codes official Anthropic features?
No. They're community-discovered conventions Claude consistently recognizes. Each one is tested before being added.
Made by Samarth at clskillshub.com. If you have a question this page doesn’t answer, email samarth@clskillshub.com.