Why Most "Claude for Founders" Advice Is Wrong
Most prompt guides for founders are written by people who haven't shipped a product. They recommend generic prompts like "act as a startup advisor" or "brainstorm 10 ideas for my business" — which sound useful and deliver nothing.
I've been running a solo B2B SaaS for six months. I've used Claude for genuinely hard founder work: pricing a new product, reframing stuck problems, writing cold emails to investors, making the build-vs-buy call on features.
These are the 10 prompt codes I actually use daily, with the specific founder situation each one solves. All tested against baseline — see how to verify Claude prompts yourself.
1. L99 — For any "should I X or Y" decision
Founder use case: Any decision you've been stuck on for more than a week.
Example: "L99 Should I raise a seed round now or bootstrap to $20K MRR first?"
What it does: Forces Claude to commit to ONE specific answer with reasoning, instead of the usual "here are the tradeoffs to consider" enumeration that doesn't actually help you decide.
Without L99: You get a paragraph about the pros and cons of each option and a non-answer like "depends on your situation."
With L99: Claude picks one, tells you why, and names the specific condition under which the other option would be correct.
This is the single most useful founder prompt. If you only learn one, learn this one.
2. /skeptic — For when you're not sure you're solving the right problem
Founder use case: "My retention is bad, how do I fix it?" type questions where you suspect you're optimizing the wrong thing.
Example: "/skeptic We're losing 30% of trial users before day 3. How do we fix activation?"
What it does: Before answering, Claude checks if your question has a wrong premise. If your trial conversion is the symptom and the real problem is elsewhere, /skeptic catches it.
What you get: "You're assuming the activation step is broken. Before I answer that, let me question it: what does your day-0 signup experience promise users? If your landing page promises faster-than-reality outcomes, you have an expectation problem, not an activation problem."
For founders, this is the code that prevents you from spending 3 months solving the wrong problem.
3. /blindspots — For pre-mortem on any plan
Founder use case: You have a plan you're about to execute. You want Claude to find the assumptions you didn't notice you were making.
Example: "/blindspots I'm about to launch a $99 monthly plan for small teams. Price it right, position it well. What am I missing?"
What it does: Surfaces the 2-3 things you probably haven't considered that change the answer.
What you get: Warnings like "you're pricing based on what competitors charge, not what your users would pay at different levels. Have you done a price sensitivity test? Also, monthly vs annual is a decision you're implicitly making — a $999/year option typically 2x's revenue from the same cohort."
4. PARETO — For synthesizing too-much-information into decisions
Founder use case: You have 5 user interviews, 3 competitor teardowns, and 10 Reddit threads on a topic. You need the 20% that actually matters.
Example: "PARETO Here are 5 user interviews about pricing [paste]. Which 2-3 insights actually change what I should charge?"
What it does: Surfaces the 20% of insights driving 80% of the decision, instead of summarizing everything equally.
5. /ghost — For any writing that can't read like AI
Founder use case: Cold emails, LinkedIn posts, outreach to VCs, customer replies. Anything where AI-detection hurts you.
Example: "/ghost Rewrite this outreach email so it doesn't sound like ChatGPT: [paste draft]"
What it does: Strips em-dashes, transitional phrases ("In today's competitive landscape"), "Furthermore," "I hope this email finds you well," and the other structural fingerprints of AI writing.
Paste your first draft. Run /ghost. Ship the result. This saves founders 20-30% of their communication time.
6. /punch — For tightening any message
Founder use case: Slack messages that are too long, pitch decks that bury the point, landing page copy that reads limp.
Example: "/punch Tighten this pitch into one sentence: [paste paragraph]"
What it does: Makes every sentence load-bearing. Removes qualifiers. Cuts filler.
Works especially well stacked with /trim: /punch /trim Rewrite this cold email to be 40% shorter without losing any content.
7. /mirror — For matching someone's writing style
Founder use case: Replying to an investor's email, writing in a co-founder's voice, matching a customer's tone before negotiating.
Example: "/mirror Read this investor's email: [paste]. Now write my reply matching their tone exactly. Reply content: [what you want to say]"
What it does: Analyzes the reference writing's formality, length, rhythm, and vocabulary — then locks Claude into matching it across the entire response.
8. OODA — For strategic decisions under uncertainty
Founder use case: "Competitor just launched a similar feature, how do we respond?" or "We have 4 months of runway, what's the play?"
Example: "OODA Competitor X just shipped [feature] at a lower price. We have 5 months runway and 80% of revenue from SMBs they now target. What do we do?"
What it does: Applies the Observe-Orient-Decide-Act framework. You get a response structured as: here's what's actually happening (Observe), here's what it means for you specifically (Orient), here's the decision (Decide), here's the first concrete step (Act).
9. PERSONA — For getting expert-level advice
Founder use case: You need to think like a specific type of expert — a B2B sales leader, a corporate development VP, an HR director hiring executives.
Example: "PERSONA Act as a VP of Sales who has built and sold 3 B2B SaaS sales teams from zero to $10M ARR. She's direct, hates fluff, thinks most sales advice is bad. Question: when should I make my first sales hire?"
Critical nuance: PERSONA only works with real specifics. "Act as a sales expert" is placebo. "Act as a VP of Sales at a Series B fintech who just sold her company for $200M" is reasoning-shift.
10. /rephrase — For making a hard message land
Founder use case: Giving critical feedback to a team member, declining a partnership request, negotiating with a customer, telling an investor no.
Example: "/rephrase Rewrite this so it's firm but not combative: [paste draft]"
What it does: Keeps the meaning exact. Changes the delivery to be appropriate for the relationship.
The Combos Founders Actually Run Daily
Single codes are fine. Stacks are where it gets powerful.
Strategic decisions: L99 /skeptic /blindspots [your question]
Cold emails: /ghost /punch /trim /voice [casual/formal] [email draft]
Investor updates: /punch /rephrase [professional but warm] [your draft]
Reframing stuck problems: /skeptic /blindspots OODA [problem description]
Hiring decisions: PERSONA [specific VP of HR profile] L99 [the decision]
Browse more task→stack combos in the Combo Generator →
The Founder Tool Built on These Codes
If you want these codes applied specifically to sales situations with 10 different founder voices (Collison, Benioff, Lütke, Chesky, Huang, Altman, Amodei, Levie, Butterfield, Lemkin) loaded as skill files — that's the Sales Agent Pack. $359 one-time, runs locally, bring-your-own Anthropic API key.
It's the same L99 / /skeptic / /blindspots logic, but applied through a specific founder's public writing as the anchor. Collison-mode gives you the Collison reframe. Benioff-mode gives you the Benioff enterprise-pricing angle. Lemkin-mode gives you the operator playbook.
Not required. These 10 codes alone get you 80% of the value.
FAQ
Which 3 should I start with if I only have 10 minutes?
L99, /ghost, /punch. L99 for decisions. /ghost for any writing. /punch for tightening. You'll use all three within 24 hours of learning them.
Do these work with Claude Haiku or only Opus/Sonnet?
All three tested cleanly on Haiku 4.5. Sonnet 4.6 produces better output on /skeptic and /blindspots (needs more reasoning depth). Use Sonnet for strategic codes, Haiku is fine for /ghost / /punch / /trim.
Is there a free version of the full library?
Yes. 100 codes free at /prompts, plus 6 free combo recipes at /combo for paste-ready templates. The full 120 with tested before/after + classification + failure modes is in the Cheat Sheet from $10.
Do you have prompts specifically for fundraising?
The fundraising-focused voice is in the Sales Agent Pack — specifically Altman-mode and Lemkin-mode are tuned for round sizing and investor conversations. The generic L99 + /skeptic combo works too, but the founder voices are sharper.
What if I'm a non-technical founder?
These codes work with Claude.ai (the web app), not just Claude Code. You don't need to be a developer. Paste them as prefixes at the start of any Claude prompt.
One Action Right Now
Open Claude.ai. Paste this:
L99 /skeptic[the single biggest business decision you've been procrastinating for the past 2 weeks]
Whatever the decision is. Paste it with that prefix. Read the answer.
That 30-second interaction will tell you more about these codes than any article I can write.