AI Guides › Workbench
By Nigel Guy · 7 min read
Ask Claude for business advice and you usually get your own view back, tidied up and politely encouraged. The advice is not wrong. It is built from the framing you supplied, so it cannot tell you that you have framed the problem badly. That is advice theatre: you leave feeling backed, and no closer to a decision.
The rule: make Claude name what kind of decision this is before it advises, and let it use only thinking that is publicly documented for that kind.
The fix is a saved skill with two moves. First it classifies your problem into one of six types and tells you if you picked the wrong one. Then it routes the problem to two or three named, published ideas that fit that type, and no others. Below is the skill, three ways to install it, and how to use it without being flattered.
| Type | The real question | Published idea it routes to | Where it comes from | Catch |
|---|---|---|---|---|
| Distribution | Does anyone know you exist? | Do things that don't scale | Paul Graham essay, "Do Things that Don't Scale" | Written about early-stage startups; weaker for a mature business |
| Product | Do they know you and not want it, or not stay? | First-principles reasoning | A general method, not one person's quote | Needs real facts from you, or it reasons from nothing |
| Focus | Too many things, none working? | Circle of competence | Warren Buffett, 1996 Berkshire Hathaway letter | Written for investing; you must translate it to operating |
| People | Hiring, firing, performance? | Nothing reliable is bundled here | None | The skill should say so and ask you for specifics, not borrow a famous name |
| Risk | Big, hard to reverse? | One-way and two-way doors | Jeff Bezos, Amazon shareholder letters | Reversibility is your judgement, not a fact the model can check |
| Money | Pricing, spend, runway? | Default alive or default dead; ramen profitable | Paul Graham essays | Needs your real numbers |
Positioning can be added as an optional seventh lens: Peter Thiel's argument that you should aim to be the only option in a niche, not a slightly better one (his 2014 Wall Street Journal piece, adapted from Zero to One). It is a view, not a law. Use it as something to argue with.
I checked the primary texts for each of these. What each source says, in brief:
A skill is a folder holding a SKILL.md file: a short header with a name and description, then instructions. Claude reads the description to decide when to load it. Save this as founder-mentor/SKILL.md.
---
name: founder-mentor
description: Classifies a business decision (distribution, product, focus, people, risk or money) before advising, then applies only published, attributable thinking that fits. Use when I am stuck on a business, hiring, pricing, or big-bet decision.
---
# Role
You are a sceptical panel of experienced operators. Agreeing with me is the lazy answer, so test it before you write it.
# Goal
A good result is one where I may change my mind about what the problem is, and I leave with one dated action and a reason I can reuse.
# Inputs I will give you
- Decision: [DECISION_IN_ONE_PARAGRAPH]
- Stage and size: [STAGE_AND_TEAM_SIZE]
- Key numbers: [REVENUE_COSTS_CASH_RUNWAY_AS_RELEVANT]
- Deadline and what is reversible: [DEADLINE_AND_WHAT_CAN_BE_UNDONE]
If any are missing, ask for them first. Do not estimate figures on my behalf.
# Steps
1. Classify. In one line, pick the type: distribution, product, focus, people, risk or money. Say whether my own wording points to a different type, and quote the phrase that shows it.
2. Choose two or three lenses that genuinely fit that type, from the list below. Drop any that do not fit.
- Distribution: doing things that do not scale, as in Paul Graham's essay.
- Product: first-principles reasoning; remove, then simplify, then automate.
- Focus: Buffett's circle of competence, and the opportunity cost of each yes.
- Risk: Bezos's one-way versus two-way doors.
- Money: Graham's default alive or dead; ramen profitable.
- Positioning: Peter Thiel's case for being the only choice in a niche.
- People: no bundled lens. Ask me for specifics and reason from them.
3. Answer in this order: (a) what this really is; (b) what each lens implies, with the reasoning, not a slogan; (c) what they would challenge, quoting my own words; (d) what is missing, such as a number, a person or a date; (e) one action I can start this week; (f) the principle behind it.
# Rules
- Attribute only views the person has published. Paraphrase; do not put words in anyone's mouth. If unsure of a source, say so.
- If the framework does not suit my scale or situation, say that plainly.
- No pep talk. No reassurance without evidence.
# Self-check before answering
Did I classify first? Did I challenge at least one thing I said? Is every attribution one I can stand behind? Is the action specific and dated?
Fill in the four bracketed inputs each time you use it. Everything else stays as saved.
| Route | Where | Cost at time of writing | Best for | Catch |
|---|---|---|---|---|
| Claude app | Customize > Skills, then "+", "+ Create skill", "Upload a skill" with the folder zipped | Skills are listed as available on Free, Pro, Max, Team and Enterprise; code execution must be on. Pro is $20 a month, about £15 at time of writing; check the £ price at checkout | Everyday use in chat | Anthropic's help page says the folder name must match the skill name. Team and Enterprise owners must enable skills first |
| Claude Code | ~/.claude/skills/founder-mentor/SKILL.md for all projects, or .claude/skills/ inside one repository |
Included with a plan that includes Claude Code; check the current plan page | Running it beside your own documents | Invoke with /founder-mentor, or let the description trigger it |
| Plain prompt | Paste the body into a chat or a Project's instructions | Free | Trying it once | You lose the automatic trigger |
If the skill fires too often, tighten the description. If it never fires, call it by name first.