One Of These Eight Founder Prompts Can Tell You To Stop. It Is Number Two.

Eight prompts for founders, posted one to a comment and numbered one through eight. One-page business plan, market validation, revenue model, go-to-market, competitor analysis, financial projection, risk assessment, investor summary. The caption promises the analytical rigour of a consulting engagement.

They are good prompts. Each one names a role, names a deliverable, and asks for something specific enough that you could tell whether the model did it — three scenarios in the financial projection, five competitors in the teardown, exactly ten first customers in the go-to-market plan. That is well above average for a pack of this kind, and it is the second strong business pack from the same page, with almost no overlap between them.

But the numbers on the comments are not decoration. They are an order of operations, and read as an order of operations the pack has a problem that has nothing to do with how any individual prompt is written. Exactly one of the eight is allowed to end the project. It is the second one. And the first one has already had you write the plan.

The eight, in the order they were given

Here they are with the role each one assigns and the question that matters: after running this prompt, could you reasonably decide not to do the business?

One exit, and it is behind the first door

Prompt 2 is the only one in the pack that ends in a decision rather than a document. Everything else produces an artefact — a plan, a model, a projection, a pitch — and an artefact, once it exists, is an argument for continuing. You do not write a twelve-month cash flow and then conclude that the business should not exist. You write it and then look for the month it turns positive.

That makes the position of prompt 2 the most consequential design choice on the card, and the card puts a document in front of it. Prompt 1 asks you to turn your idea into a one-page plan with a ninety-day roadmap. By the time you run prompt 2, you have a roadmap. The ruthless analyst is now arguing with something you already own.

This is not a theoretical worry about prompt ordering. It is the oldest failure mode in starting anything: the plan is cheap to make and expensive to abandon, and every hour you put into it raises the price of walking away. A prompt pack that generates the plan first and the verdict second has built that in.

And look at how the card hands out roles, because the author clearly knew which prompt carried the weight. Seven of the eight roles are graded by seniority or skill — top-tier, expert, savvy, experienced, senior, startup CEO twice. Exactly one is graded by disposition: an honest and ruthless market analyst. The adjective changes from how good they are to how willing they are to tell you something you do not want to hear. That is the right instinct in the wrong slot.

The same eight prompts, in an order that can still stop you

Nothing here needs rewriting. The fix is which comment you open first.

The reordering costs nothing. It is the same eight prompts, the same eight comments, read in a different sequence. What changes is that the three prompts capable of producing bad news — the verdict, the competitor who is already winning, the risk that must be fixed before launch — all run while the project is still free to cancel.

The second gate, which the card almost hides

Prompt 7 is the risk assessment, and its last clause is easy to skim past: rank the risks by urgency and determine which must be addressed prior to launch. That is a gate. It is the one prompt besides number 2 that can return a sentence you are obliged to act on before continuing.

But it arrives at position seven, after the revenue model, the go-to-market plan, the competitor teardown and the twelve-month projection. By then the founder has five documents and a pitch half-written. The risk report is being read by someone who has already decided.

It is also the better-built of the two gates. Prompt 2 asks for a verdict but does not say what evidence the verdict should rest on. Prompt 7 names five risk categories — market, execution, finance, legal and compliance, competition — asks for likelihood and impact on each, asks for early warning signs, and asks for a low-cost test. That is a structure you could actually argue with. Run it second and prompt 2 gets a lot more useful, because the verdict has something underneath it.

“Claude can now” — what actually changed, and it is not a prompt

The caption opens with a siren and the word BREAKING, and says Claude can now approach business strategy with consulting-grade rigour. Nothing in the eight prompts needs a new capability. Naming a role and asking for a structured deliverable has worked on every general-purpose model for years, and it would have worked on these eight in 2024.

There is a real answer to what changed for founders, though, and it is a few clicks away. Anthropic keeps a public repository of free plugins for knowledge work. It was created on 23 January 2026, it was last pushed on 23 October 2026 — the same day this card went up — and it carries over 32,000 stars. Inside it, the small-business plugin alone ships 44 skills, with names like cash-flow-snapshot, business-pulse, growth-pulse, report-builder, proposal-builder, lead-finder and tax-prep. The separate finance plugin adds eight more, including financial-statements and variance-analysis.

The difference between those and these eight prompts is not quality. It is persistence. A prompt is something you paste again every time; a skill is a file that stays installed and can be improved when it gets something wrong. If you are going to run prompt 6 every month for a year, the twelfth run should be better than the first, and a comment on Facebook has no way to make that happen. We went through that repository in detail when another card undercounted it.

What the pack gets right

The ordering problem is structural, not a writing problem. Line by line these are better than most.

Four honest caveats

1. The numbering may not be an instruction. The prompts are numbered one to eight and Facebook does not display comments in that order anyway, so a reader may well shuffle them without being told to. My argument is that the numbers are the only sequencing signal the pack gives, and that the signal points the wrong way. If the author intended them as a menu rather than a method, the fix is a sentence saying so.
2. I have not run them. Whether prompt 2 actually returns “walk away” on a weak idea, or softens into “pivot” for almost everything, is an empirical question. Models are agreeable by default, and “honest and ruthless” in a role line is a weaker instrument than a worked example of a rejection. Test it on an idea you already know is bad before you trust it on one you like.
3. “Walk away” from a model is advice, not a finding. None of these prompts gives the model access to your market, your numbers or your customers. A verdict produced from a paragraph of description is a prompt for your own thinking, not a substitute for it, and no reordering changes that.
4. The plugin comparison is about durability, not quality. I have not tested the 44 small-business skills against these eight prompts and I am not claiming the skills produce better output. The claim is narrower: a file you can edit accumulates improvements and a comment you re-paste does not.

The one-line version

Open comment two first, then five, then seven — the validation verdict, the competitor teardown and the ranked risks — and only then write the plan, the pricing and the projection, because the one prompt in this pack that is allowed to tell you to stop is currently sitting behind the one that makes stopping expensive.