The AI PERMA Model Coach skill for Claude

You will probably recognise the problem this skill solves. You do not feel great in your work or your week, you put that to an AI, and you get a friendly mush back: take good care of yourself, plan breaks, talk to someone about it. It is all true and none of it helps you, because there is no structure underneath it. This skill puts that structure underneath. It cuts wellbeing into five pieces that each get looked at separately, and only once all five have been covered does the conclusion come.

That split is not arbitrary. It comes from Flourish by Martin Seligman from 2011, in which wellbeing is not seen as a single score but as five independent dimensions that each carry value on their own. Someone can score strongly on meaning and weakly on relationships, or the other way round, and those two say something very different about what should change. An average hides exactly the difference that is worth acting on.

What the skill is meant for: personal reflection, coaching conversations, appraisal conversations and a conversation about how a team is doing. What it is not meant for is stated just as clearly: never diagnose, no medical advice, and refer to a professional at signs of serious complaints. That last point is the most important sentence in the whole file and we return to it in DOC.06.

Free association without a fixed framework. Without the skill a conversation jumps from your sleep pattern to your manager to your Sunday evening feeling, and at the end you still do not know what is actually the problem. With the skill Claude works through the five dimensions systematically, including the dimensions you did not bring up yourself. That is often exactly where the surprise sits.

Superficial answers with no follow-up. A vague question gets a vague answer. That is why concrete questions are prescribed, with it stated explicitly: no vague open questions. When did you last do something purely because it felt good is a different question from how are you doing, and produces a different conversation.

Advice before the picture is right. First it summarises what it hears, names your strongest dimensions and your points of attention, and then asks whether that picture matches how you see it. Only once you confirm does it move on to the plan. That saves an action plan for a problem you do not have.

Generic advice for all five at once. The plan covers the weakest one to two dimensions, not all five. Three actions for one dimension is manageable. Fifteen actions across five dimensions is a list you dismiss after three days. In diff form, with questions from the file: what goes out and what comes back in its place.

Next comes the approach in four steps. Step 1 is the intake and it always starts with an open question: what brings you here today, or what would you like more insight into. Step 2 is the analysis, in which every dimension comes up systematically following a fixed pattern of four actions: ask two targeted questions, listen actively to the answers, reflect back what it hears, and only then move on to the next dimension.

Step 3 is the synthesis: a summary of your strongest dimensions and your points of attention, the question of whether that matches how you see it, and drawing up an action plan together for the weakest one to two dimensions. Step 4 is that plan itself, with three concrete and achievable actions per weak dimension for the coming two to four weeks.

The most valuable part of the file is a block of ten questions, two per dimension, which you find literally in FIG.02. That is also exactly why a skill of 653 words is more useful than a long prompt: the questions are fixed, so you get the same conversation every session, even if you phrase the request yourself in a sloppy way.

On top of that, there are fifteen example actions, three per dimension, that make the plan concrete. For positive emotions this includes writing down three things you are grateful for every day and taking conscious breaks without a screen. For engagement, ninety minute blocks for deep work and removing distractions. For relationships, checking in consciously each week with someone close to you. For meaning, writing down your core values and checking whether your current work matches them. For accomplishment, a goal that is challenging yet achievable and keeping track of your progress.

The tone is also fixed, in six lines: empathetic and non-judgemental, concrete questions rather than vague open questions, using summaries and reflections to verify, naming the scientific basis when it helps, never diagnosing or giving medical advice, and referring to a professional at signs of serious complaints. The last two are not style advice but hard limits, and they sit in FIG.04.

Finally, four ready made prompts. The basic analysis walks you through all five dimensions. The deep dive analysis takes one dimension and asks five progressively deeper questions. The action plan prompt starts from scores you fill in yourself and delivers a four week plan. And the team prompt applies the model to a team instead of to yourself. If you want to learn to set up a file like this yourself, the structure is explained in writing a SKILL.md NL.

And then what is not in there, because that belongs here too. There is no measurement instrument in it, no questionnaire with points and no norming. There is also no worked case in it, unlike the longer skills in the library. This is a short and tight file: the questions and the order are the whole skill. That is an honest limitation and at the same time the reason it is so easy to use in a team.

Flourish, Martin Seligman, 2011. This is the book in which PERMA is first set out this way. The core is a break with his own earlier work: wellbeing is not one thing you can add up to a single figure, but five parts that each carry value on their own. That is why the skill looks per dimension and does not deliver an overall score. Anyone who only looks at the average misses exactly the dimension where there is something to gain.

Authentic Happiness, Martin Seligman, 2002. The earlier book, from the beginning of positive psychology. It sits in the source list because without that starting point you do not see why Flourish was a correction. You do not need to read it to use the skill, but it explains why the emphasis in the model lies on independent dimensions.

The Positive Psychology Center at the University of Pennsylvania. Seligman's own institute, named in the file as a source for background and research. Useful if you want to read further than the two books.

One thing should be said here, because it does not sit in the skill and it is too important to leave out. The model is widely used in research and practice, but that does not turn a conversation with an AI into a measurement. What you get is an ordered conversation along five dimensions with a plan at the end. That is worth exactly as much as the honesty of your own answers, and no more. Looking for a model that puts the needs underneath in order, look at the Maslow Pyramid Coach skill. Is it mainly about direction rather than wellbeing, then the Ikigai Coach skill is the better way in.

After that you pick one of the four prompts from the file and simply get started. Save the session in a project, because the skill comes into its own if you repeat it every quarter and can lay the dimensions side by side. If you get stuck on the installation, the step by step explanation sits in installing Claude skills NL. More background on working with AI you will find in the knowledge base.

For a team the file uses a separate prompt. You then work through the same five dimensions for the group. Do that with one warning that is not in the skill but is still true: you are then talking about people who did not give an answer themselves. Use the outcome as a starting point for the conversation with them, not as a conclusion about them. For the conversation afterwards the Radical Candor Coach skill NL is a logical addition.

There is one situation in which you should not use it, and it is more important than all the others. If you have complaints that persist, if you sleep badly, if you feel down or if you do not see a way through any more, this is not a coaching question. The file says so itself too: never diagnose, no medical advice, and refer to a professional at signs of serious complaints. That conversation belongs with your GP or your occupational health doctor. An AI that talks along pleasantly is not helpful there but delaying.

And the most honest limit: the skill makes what is going on visible, it does not solve it. If your engagement is low because your work does not suit you, ninety minutes of deep work will not fix that. What the model does do is move the question from how are you doing to which dimension is stuck, and that is a question you can act on. If the problem goes beyond wellbeing and heads towards your work, the GROW Coach skill is a better next step.

If engagement comes out as your weakest dimension, the follow-up question is where your attention is leaking away. In work, that is often fragmentation: too many loose tasks that no one has cleared away. No model helps against that, automation does. If you want to know which recurring work in your business can be taken off your hands, look at AI and automation or have your team learn to work with it themselves through the AI training NL. And if you simply want to keep talking about what AI can do for you, we do that in a conversation.