Prompts11 min read

The AI Codes I Actually Use to Build, Ship, and Not Get Hacked

Most "AI prompt code" lists are the same five tricks with different names.

These are the ones I actually use.

Some started from problems I ran into while building my own products. Others are just shortcuts for prompts I kept typing over and over. I put them all here so I don't have to keep digging through old chats to find them.

119 so far. The ones at the top are the codes I use most. The rest are in the library.

The code itself isn't magic. It's just a shortcut for telling the AI what job you want it to do. You can use the code on its own when the meaning is obvious, or copy the prompt underneath it when you want the full instruction.

Build & Ship

/KILL — Try to kill the idea before I waste time building it.
Prompt · copy & paste
Act as a skeptical critic of this idea: [describe it]. Who already solves this, how well, and what would make someone switch? Then give me the three most likely reasons this fails in 6 months. Don't soften it.
/SHIP — Build the smallest version that proves the main thing works.
Prompt · copy & paste
Build the smallest possible version of [idea] that proves the core flow works end to end. This is a prototype, not a production launch. Skip anything that isn't needed to test the main idea today.
/BORING — Let AI handle the stuff that isn't the reason I'm building the product.
Prompt · copy & paste
Add [auth / schema / integration] to this app. Before writing anything, tell me what you're going to change, why, and where you're making assumptions. Don't touch unrelated parts of the codebase.
/MVP — Turn a vague feature into something I can actually build.
Prompt · copy & paste
Turn this vague feature request into a concrete implementation spec: [paste request]. Define the intended behavior, inputs, outputs, constraints, edge cases, and what "done" means.
/SCOPE — Catch the stuff I'm about to build that nobody needs yet.
Prompt · copy & paste
Look at what I'm about to build: [describe it]. What's the smallest version that still proves the point? What am I adding that a first user doesn't need? Cut anything that isn't necessary to test the core idea.
/POSTLAUNCH — Find the thing I'm overlooking after launch.
Prompt · copy & paste
Here's what changed this week: [list]. Here's what users did that surprised me: [list]. Based on that, what is the highest-risk thing in the product that I still haven't tested?

Security & Production

/BREAKIT — Try to break the app instead of politely reviewing it.
Prompt · copy & paste
Act like a security engineer trying to break this app. Look for concrete attack paths through authentication, authorization, sessions, payments, user data, uploads, secrets, injection, SSRF, and rate limits. Rank the findings by impact and exploitability, then show me the fix for each.
/COOKIE — Check whether someone can forge or reuse a session.
Prompt · copy & paste
Trace how sessions and cookies are created, signed, stored, refreshed, and verified. Look for ways a user could forge, replay, modify, or reuse a session to access another account.
/LEAK — Find places where one user could reach another user's data.
Prompt · copy & paste
Trace every user-controlled ID through the API routes, database queries, storage paths, and authorization checks. Look for any way one user could read, change, or delete another user's data.
/WOULDBREAK — Ask what I'm missing before I trust the code.
Prompt · copy & paste
What's the most likely thing about this implementation that would work in testing but fail in production? Explain why, what would trigger it, and how I'd know it happened.
/HEAL — Recover from transient failures without hiding repeated failures.
Prompt · copy & paste
Add bounded retries, failure logging, alerting, and a visible last-success/last-failure status to this background job. Transient failures should retry automatically, but repeated failure should become impossible to miss.
/RATELIMIT — Find where someone could run up my bill.
Prompt · copy & paste
Look at every expensive endpoint in this app, including AI calls, searches, emails, file processing, and external APIs. What stops someone from abusing each one? Show me the current protection, the likely cost exposure, and what I'd change.

Money & Business

/BILL — Make billing something I never have to babysit.
Prompt · copy & paste
Set up Stripe so a customer can subscribe, upgrade, downgrade, cancel, and request a refund without needing me. Keep subscription state in sync with webhook events, and handle duplicate, delayed, failed, and out-of-order events safely.
/WEBHOOK — Find the ways a webhook can quietly go wrong.
Prompt · copy & paste
Review this webhook handler for duplicate delivery, out-of-order delivery, delayed delivery, missed delivery, retries, invalid signatures, partial failures, and database inconsistencies. Tell me which cases are handled and which aren't.
/PRICE — Stress-test a price point from both sides.
Prompt · copy & paste
Here's what it costs me to run this product monthly: [figure]. Here's what I'm charging: [figure]. Stress-test this price from both sides. Would a customer see enough value to justify it, and does the margin leave me enough room to grow? Tell me what I'm underestimating.
/CHURN — Find the actual reasons someone would leave, not just the cancel flow.
Prompt · copy & paste
Look at this product, its onboarding, usage, pricing, and cancellation flow. What are the most likely reasons a customer would leave? Separate problems I can actually fix from reasons I probably can't control.
/COLDSTART — Find the first 10 users without pretending I have an audience.
Prompt · copy & paste
I'm launching [product] with zero audience. Where are people already talking about the exact problem this solves? Give me specific communities, what they're already discussing, and a useful way to show up without spamming them.

Thinking & Decisions

/ULTRA — Look past the obvious answer.
Prompt · copy & paste
Analyze this carefully before answering. Check the obvious answer first, then look for hidden assumptions, important edge cases, non-obvious alternatives, and second-order effects. Prioritize correctness and useful insight over length.
/REDTEAM — Try to change my mind.
Prompt · copy & paste
Take the strongest position against what I just said. Find the biggest weaknesses, failure modes, and assumptions I'm relying on. Don't agree with me just because the idea sounds reasonable.
/EVIDENCE — Separate what we know from what we're assuming.
Prompt · copy & paste
Go through the important claims in your answer and label each one as verified fact, inference, estimate, assumption, or unknown. For anything that needs verification, tell me what evidence would settle it.
/3X — Give me three genuinely different ways to do it.
Prompt · copy & paste
Give me three substantially different approaches to this problem. Don't give me three rewrites of the same idea. Make the tradeoffs between them clear.
/RISK — Find the one assumption that matters most.
Prompt · copy & paste
Out of everything in this plan, what's the single highest-risk assumption? Explain why it matters, what failure would look like, and what I could test quickly to reduce the uncertainty.

Writing & Voice

/HUMAN — Make it sound like a person wrote it.
Prompt · copy & paste
Rewrite this so it sounds like a sharp, natural person wrote it. Cut corporate filler, generic transitions, fake enthusiasm, repetitive sentence structures, and anything that sounds like it's trying to sound smart.
/CUT — Make it half as long without making it worse.
Prompt · copy & paste
Cut this to roughly half the length. Keep the important information and specific details. Remove setup, repetition, hedging, and anything the reader doesn't need.
/NOFLUFF — Remove anything that isn't doing work.
Prompt · copy & paste
Go through this sentence by sentence. Cut anything that doesn't add new information, clarify the idea, or move the reader forward. Keep the useful details.
/VOICE — Match how a person actually talks.
Prompt · copy & paste
Rewrite this using the tone of the example below: [paste example]. Match the sentence length, directness, rhythm, and level of casualness. Copy the feel, not the wording.
/TIGHTEN — Find the weakest sentence.
Prompt · copy & paste
Read this and find the single weakest sentence. Tell me why it's weak, then give me a stronger replacement that fits the surrounding paragraph.

Build & Ship

The rest are shorter by design. Find the code you need, paste the prompt, and keep moving.

/SPEC — Turn a vague request into an unambiguous build spec.
/FIRSTPRINCIPLES — Strip the problem down to fundamental facts and rebuild the solution from there.
/DECOMPOSE — Break the task into smaller problems and solve them one at a time.
/PLANFIRST — Work out the approach before touching the implementation.
/REVERSE — Start with the desired outcome and work backward to what must happen first.
/INVERSION — Ask what would produce the opposite result and use that to expose weaknesses.
/SECONDORDER — Look at the downstream effects, not just the immediate result.
/REFRAME — Look at the same problem from several fundamentally different angles.
/SYSTEMMAP — Map the people, incentives, dependencies, and feedback loops around the problem.
/TESTFIRST — Define the important tests and edge cases before writing the implementation.
/MINIMALCHANGE — Fix the problem with the smallest safe change and leave unrelated code alone.
/ROADMAP — Turn the strategy into an ordered implementation plan.
/PRIORITY — Separate what must happen from what can wait.
/EXECUTE — Stop analyzing and turn the conclusion into concrete next actions.
/AUTOMATE — Find the repetitive work that can be automated or systematized.

Security & Production

/THREATMODEL — Map the assets, trust boundaries, attackers, and attack surfaces.
/SELFTEST — Check the implementation against its requirements before calling it done.
/SECONDLOOK — Revisit the first conclusion and look for what got missed.
/FAILUREMODES — List concrete ways the system or plan could fail.
/PREMORTEM — Assume the plan failed and work backward to why.
/REDFLAGS — Find warning signs that deserve more investigation.
/PUSHBACK — Challenge assumptions before moving forward.
/SKEPTIC — Don't accept the premise until it survives scrutiny.
/STEELMAN — Build the strongest reasonable version of the opposing argument.
/FALSIFY — Look specifically for evidence that could prove the conclusion wrong.
/COUNTER — Give the strongest counterargument to the conclusion.
/VERIFY — Verify important claims before relying on them.
/FACTCHECK — Check factual claims and flag unsupported ones.
/PRIMARY — Prefer primary or authoritative sources.
/SOURCECHECK — Make sure the source actually supports the claim.
/TRIANGULATE — Check important conclusions against independent evidence.
/CONTRADICT — Look for credible evidence that conflicts with the leading conclusion.
/UNCERTAINTY — Make important unknowns explicit.
/CLAIMMAP — Map major claims to the evidence supporting them.
/RECENCY — Separate current information from information that may now be outdated.

Decisions & Tradeoffs

/ASSUMPTIONS — List every important assumption the answer depends on.
/TRADEOFFS — Show what each option gains and gives up.
/COMPARE — Compare options using the same criteria.
/CRITERIA — Figure out what actually matters before choosing.
/WEIGHTS — Give more weight to the criteria that matter most.
/PARETO — Find the small number of factors driving most of the result.
/SENSITIVITY — Find the assumptions that would most change the conclusion.
/REVERSIBILITY — Separate easy-to-undo decisions from expensive ones.
/OPPORTUNITYCOST — Account for what choosing one option means giving up elsewhere.
/VALUEOFINFO — Find the missing information most worth getting before deciding.

Scenarios & Uncertainty

/HYPOTHESES — Generate several plausible explanations before picking one.
/SCENARIOS — Show how the answer changes across different plausible futures.
/BASERATE — Compare the situation with relevant historical frequencies or benchmarks.
/BESTCASE — Show the realistic upside case.
/WORSTCASE — Show the realistic downside case.
/ROBUST — Find an answer that still works when assumptions change.
/THRESHOLD — Identify what evidence would actually change the decision.
/RANGE — Give a plausible range instead of fake precision.
/EDGECASES — Look for unusual cases that could still matter a lot.
/PROBABILISTIC — Make uncertainty explicit instead of pretending to know more than we do.

Creativity

/BRAINSTORM — Generate many possibilities before filtering them.
/REMIX — Keep the useful parts but rebuild the structure.
/MASHUP — Combine mechanisms from different ideas or industries.
/SCAMPER — Systematically substitute, combine, adapt, modify, remove, or reverse elements.
/COUNTERFACTUAL — See what changes when a key assumption is reversed.
/WILD — Generate the bold ideas first, then filter for practicality.
/MINIMAL — Strip the solution down to the smallest version worth testing.
/ANALOGY — Borrow a useful mechanism from another domain.
/CONSTRAINT — Add a hard constraint to force different solutions.

Learning

/EL10 — Explain it like you're 10.
/PLAIN — Remove jargon and just say the thing.
/TEACHME — Teach it progressively instead of dumping everything at once.
/SOCRATIC — Teach through questions that lead toward understanding.
/EXAMPLES — Use concrete examples to make the idea click.
/METAPHOR — Explain the concept using an intuitive analogy.
/LAYERED — Start simple and add complexity only when needed.
/PREREQS — Tell me what I need to understand first.
/GAPS — Find what I'm probably missing or misunderstanding.
/QUIZ — Test whether I actually understand it.

Writing

/DIRECT — Get to the point immediately.
/SPECIFIC — Replace vague language with concrete details.
/HOOK — Make the opening earn the next sentence.
/PUNCH — Increase clarity, force, and information density.
/SPOKEN — Make it sound natural out loud.
/POLISH — Final pass for clarity, rhythm, grammar, and precision.

Business

/FOUNDER — Look at the problem like an operator who has to actually make it work.
/MARKET — Analyze customers, competitors, alternatives, and market structure.
/MOAT — Focus on what could make the product hard to copy or replace.
/CUSTOMER — Evaluate the idea from the customer's side.
/MONETIZE — Find practical ways this could make money.
/UNITECONOMICS — Look at acquisition, retention, pricing, and margin together.

Quality Control

/RUBRIC — Define what a good answer actually needs to accomplish.
/INDEPENDENT — Produce a fresh analysis before comparing it to the first one.
/FINALCHECK — Check completeness, accuracy, and instruction-following before finishing.
/OUTLINE — Structure the answer before expanding it.
/ITERATE — Draft, critique, improve, then produce the final version.
/STOPRULE — Tell the AI when enough research or iteration is enough, instead of letting it circle forever.

A note about the word "codes"

The code isn't a secret ChatGPT feature.

/HUMAN doesn't switch on a hidden "human mode." /ULTRA doesn't unlock extra intelligence. /3X isn't a built-in command.

They're just memorable shortcuts for useful instructions. That's the point. Use the code because it's easy to remember — the prompt underneath is what actually does the work.

Stack them

You don't have to use one at a time.

Prompt · copy & paste