The AI Antifragile Thinker skill for Claude

The problem this skill tackles is invisibility. Most plans look fine as long as the world behaves the way the spreadsheet assumes. The fragility sits in what is not written down: the client who is 70 per cent of revenue, the supplier with no alternative, the fixed costs that keep running if revenue drops by thirty per cent. You usually only see those weak spots after the first shock, and by then it is too late to do anything about them. This file pulls them out beforehand, and at the same time looks for the other side: the hidden options with a small downside and a large upside that stay unused in almost every plan.

This skill is written for entrepreneurs and professionals with real decisions to make: a growth plan, an investment, a pivot, a career move, an income structure. The frontmatter states explicitly who it is especially meant for: anyone who takes too much revenue from one client, depends on one channel or platform, carries large fixed costs or draws their entire income from a single source. It is part of the skill library NL that we make available for free through the AI and automation service, with no account and no sales email afterwards.

One thing up front: the skill is not an oracle and does not try to be. It does not predict shocks and nowhere promises that your plan will become invulnerable. What it does is make the shape of your risk visible, so that you choose which fragility you accept and which you cut away.

The first blind spot is concentration. Revenue that leans mostly on one or two clients feels like success, but it has a concave payoff: it can go a little better and a lot worse. In the example from the file that is the core of it: eighty per cent of revenue from two clients means that one cancellation costs forty per cent of your income. The file calls such points single points of failure, and they sit at the top of the scan: identify them, remove them or duplicate them.

The second blind spot is the assumption that has to hold. Every plan contains assumptions that are not flagged as assumptions. That a team of four will sell itself. That the market keeps behaving the way it does this quarter. So it asks explicitly: which assumptions have to HOLD for this plan not to collapse, and which implicit insurance was never bought?

The third blind spot is the missed upside. Antifragile thinking is not only defence. It looks just as hard for what actually becomes more valuable in chaos: the knowledge that grows scarcer with every market shift, the small experiment with a capped cost and an uncapped return. Anyone who only hedges risk becomes, at best, robust. The difference sits in the payoff, and here is how that difference sounds in plain language:

Then the file defines the persona: a thinker trained on the work of Taleb, from Fooled by Randomness in 2001 to Skin in the Game in 2018. Sharp, no sales talk, not pessimistic but realistic about volatility. Then come the Triad and eight core concepts that the skill applies consistently: Black Swan, convexity and concavity, the barbell strategy, via negativa, optionality, skin in the game, the Lindy effect and the single point of failure.

The core is the fixed delivery structure: eight parts in a fixed order, of which the last is optional. First what you have, in two to three sentences without flattery. Then the Triad classification per core part, each time with the shape of the payoff attached. Then the most important work: the hidden fragility, followed by the hidden antifragility. After that the actions, in a deliberate order: first via negativa (three to five things that have to go), then only afterwards the barbell with optionality. Part seven is the one sentence: the most important asymmetry plus what has to happen this week. And the Lindy check closes it off: what in the plan leans on something that has already existed for twenty years, and what is new and untested.

It is striking how many rules are about tone. No flattery, no consultancy speak, no jargon without explanation. Be direct about the downside and do not soften it. Never end on a note of success, but on the one sentence. And the file contains a separate limits section in which the skill puts its own framework in perspective: antifragility is a direction and not a ruler, the 90 to 10 split is a rule of thumb and not a law, and the Lindy effect works the other way round for people, companies and machines. Want to learn to build this kind of instruction file yourself? The full explanation is in writing a SKILL.md NL.

Around that sit three sources that lay the foundation. Mandelbrot's The (Mis)Behavior of Markets from 2004 describes the fat tails that Black Swan thinking rests on: extreme outcomes occur far more often than the normal distribution promises. The essay The Black Swan of Cairo by Taleb and Blyth from 2011 shows that suppressing volatility actually increases fragility: whoever smooths away every small shock is saving up for a big one. And Searching for Safety by Wildavsky from 1988 sets anticipation against resilience as two different safety strategies: trying to predict every hit, or making sure you can take hits.

It is useful to see what this skill is and is nót. It is not a risk analysis that puts probabilities and consequences into a matrix, and not a pre-mortem that lets one specific plan fail. For the first there is the Second Order Thinker skill, which attacks your plan as a red team, with opponents and chains of consequence. For the second there is the Pre-Mortem Analyst skill NL. The Antifragile Thinker asks a different question: not what can go wrong, but what shape your payoff has if sómething goes wrong, whatever that is.

Even without installing the skill you can use these principles to improve your plan: count how much of your revenue leans on one client, look for the assumption that hás to hold, and ask yourself for every investment whether the payoff is concave or convex.

After that, using it is simple: describe your plan or your situation in plain language and ask for a scan. The more concrete the amounts and ratios, the sharper the result; the skill asks itself for whatever is missing. If you get stuck anywhere while installing, every step is described in more detail in installing Claude skills NL, and the broader explanation of working with AI is in the knowledge base.

There are also situations where you are better off leaving it aside, and the file names them itself. If there is already a concrete plan and you want to know why it will have failed in a year, take the Pre-Mortem Analyst NL. If you want a full attack on your plan, with opponents, their motives and the chains of consequence of your actions, the Second Order Thinker is the right tool. And if it is not about your plan but about yourself, about how you stay standing under setbacks, you are better off with the Stoic Coach NL: the same awareness that the world is unpredictable, but aimed at what is in your own power.

And the most honest limit comes from the file's own limits section: antifragility is a direction, not a ruler. There is no score, no index and no benchmark, because convexity is easy to recognise in hindsight and hard to see in advance. What you get is a classification per part and per type of shock, with the reasoning attached, so that you can make the judgement call. It does not make your plan invulnerable. It never promises that either. It makes sure you know whére it is vulnerable and what that vulnerability earns or costs you.

The most common fragility at Dutch businesses is, incidentally, one this skill would flag immediately: all enquiries from one channel. Anyone who leans entirely on word of mouth or on paid ads has a single point of failure in their pipeline. The free SEO scan shows within seconds where your site stands, and that is a first step towards a second channel that cannot be taken away from you. In the guide eight decision models compared NL you can see how this model relates to the others. And if you want to talk further about what AI can do for your business beyond this, from separate skills to full automation, we simply do that in a conversation.