Thinking & Reasoning
Commands · 22 commands
/chainNewCommandsA prefix that makes Claude split any large task into a step-by-step chain of reasoning before giving its answer.
/chain [your task] Example: /chain Create a full content strategy for a startup selling B2B SaaS
/debateNewCommandsA prefix that has Claude argue both sides of a decision you are facing, so you notice what you might be missing.
/debate [your decision] Example: /debate Should I leave my 9-5 job to work full time on my side business
/devilsadvocateNewCommandsDeliberately takes the other side to expose weak assumptions and make your decisions sturdier.
/devilsadvocate [your decision or view]
/falsifyNewCommandsMakes Claude try hard to prove its own answer wrong before showing it, cutting down errors and blind spots.
/falsify [your question]
/firstprinciplesNewCommandsReduces a complex problem to its basic truths, then builds the solution back up from there.
/firstprinciples [your problem]
/godmodeNewCommandsAsks Claude for the most thorough answer it can give, dealing with every side of the question in a single reply.
/godmode [your question]
/inversionNewCommandsWorks the problem in reverse: identifies what would make failure certain, then steers away from those actions.
/inversion [your problem]
/L99NewCommandsPushes Claude to answer at its top expert level, with no simplification and nothing watered down.
/L99 [your question]
/OODANewCommandsTakes the problem through the Observe, Orient, Decide, Act cycle, a quick loop for making decisions under pressure.
/OODA [your problem]
/premortemNewCommandsTreats your plan as if it had already failed, then reasons backwards to find what caused the failure.
/premortem [your plan]
/redteamNewCommandsAttacks your idea from all sides to surface weaknesses, risks and hidden flaws before you act on it.
/redteam [your idea]
/roastNewCommandsA prefix that has Claude pull your idea apart and show every weak spot, so you see the problems before your audience does.
/roast [your idea] Example: /roast My startup idea is an AI tool that drafts cold emails on behalf of agencies
/skepticNewCommandsHas Claude question your question first, rather than accepting it as stated and answering straight away.
/skeptic [your question]
/steelmanNewCommandsRebuilds an argument in its strongest possible form before judging or criticizing it.
/steelman [the argument]
/treeNewCommandsHas Claude lay out several possible paths and only then pick the best answer, rather than settling on its first idea.
/tree [your question]
/UDANewCommandsRuns a root cause analysis in a military style, taking the problem apart from every direction.
/UDA [your problem]
/ultrathinkNewCommandsMakes Claude think the question through at full depth before replying, so the answer is reasoned out rather than a quick guess.
/ultrathink [your question]
/checkmeNewCommandsStress-tests a plan, idea, decision or argument before you commit: every flaw in your reasoning, the blind spots, the risks you are ignoring and the thing most likely to go wrong, without softening.
/checkme [your plan, idea, decision or argument] List everything that is wrong with my reasoning: the blind spots, the risks I am not looking at, and what is most likely to go wrong. No niceness. I need to hear this before I commit.
/compareNewCommandsCompares two options with honest pros and cons for each, without choosing a winner unless asked, and adds what most people miss about each one and what you would regret if you picked wrong.
/compare [option A] vs [option B] List the honest pros and cons of each option. Do not name a winner unless I ask for one. For each option, add the point most people miss and what I would regret if I made the wrong choice.
/negotiateNewCommandsWrites a word-for-word script for a hard money conversation (salary, rent, a refund, a contract or a bill): the opening, a reply to each objection, your strongest point and your walk-away line, in your own voice.
/negotiate [the situation, e.g. salary, rent, a refund, a contract, a bill] Give me the script, word for word: how to open, how to answer each objection, my strongest argument, and the limit I must never go below. It should sound like me, not like a lawyer.
/roastNewCommandsGives a blunt critique of an idea, plan, resume, piece of writing or business, with no compliments: what is weak, what would fail and what a sharper person would change right away.
/roast [your idea, plan, resume, text or business] Tell me, with total honesty and no compliments, what is wrong with this. Point out the weak spots, what would fail, and what a person smarter than me would change straight away.
/scanitNewCommandsReads a contract, terms, policy or listing line by line and points out each red flag, hidden fee, automatic renewal and trap, explained in plain English, marking the ones that could actually cost you money.
/scanit [contract, terms, policy or listing to check] Go through it line by line and list every red flag, hidden fee, automatic renewal and trap. Explain each in plain English and mark which of them could really cost me money.