The AI Atomic Habits Coach skill for Claude

The AI Atomic Habits Coach is a free skill from the skill library of TheSEO. A skill is an instruction file, SKILL.md, that gives an AI assistant a fixed method for one kind of question. Not software you install, no subscription, no link to an app: one text file of 1,590 words that tells Claude how to apply the habit framework of James Clear to your situation, in which order it does that, and where it has to stop. You download the zip at the top of this page, put the folder in your AI environment, and from that moment on Claude answers habit questions according to this logic instead of with the first motivational text that comes to hand.

The problem the skill solves is familiar enough. You want to get up earlier, scroll less, exercise more often or finally finish that book, and you start with plenty of energy. Two weeks later it is gone, and the conclusion you draw is about yourself: apparently I have too little discipline. The framework turns that round. You are not broken, your system is broken. The cue is invisible, the friction is too high, the reward comes too late, and nobody warned you about the phase in which the work is being done but there is still nothing to see.

The skill is meant for personal use: someone who wants to build or drop a habit and would rather have a system for it than encouragement. The frontmatter itself names the openings on which Claude picks it up, and that is a broad list: from "I want to build a habit", "how do I stop doing X" and "habit stacking" to sighs such as "I cannot keep going", "I fall back into old patterns", "stop scrolling", "get up earlier" and "deep work routine". Say something like that to Claude while the skill is loaded and you get it automatically, without having to remember a command.

It belongs to the skill library of 100 free skills that we make available from the AI and automation service, with no account and no sales email afterwards. If you are not sure yet what a skill is, start at what Claude skills are. Want to write such a file yourself, then writing a SKILL.md explains the structure you see in this very file.

One thing up front, because the skill says it too: this is not therapy and not a magic potion. It calls Atomic Habits literally an operational framework leaning on behavioural psychology, and it forbids Claude to turn it into pseudo-science. If there is a deeper problem underneath, it does not build a habit plan that tidies the symptom away, but points you on.

It states a position at the top and holds on to it for the rest of the file: goals lose to systems, and identity comes before behaviour. Read that as a working instruction and not as a slogan, because it literally decides the first question you get. Not: what do you want to achieve. But: who do you want to become.

The difference is in the sentence you say to yourself. The example the file itself uses is that of someone quitting smoking. "No thanks, I do not smoke" is an identity claim. "No thanks, I am trying to quit" is a goal claim, and that second sentence leaves the behaviour hanging in mid-air: you are still a smoker gritting their teeth. In the output of the skill the identity claim therefore sits at the top, in one sentence, in the first person and in the present tense. Everything below it has to fit.

A goal is also an end point, and a habit is not. Anyone who loses weight to hit a goal falls back afterwards into the behaviour that put the weight on. Someone who sees themselves as a person who respects their own body has no end point at all, only a direction. So it asks explicitly for wording at that level, and not for figures at all.

The second reason systems win: scale. The framework counts on small, repeated improvements, the familiar 1 per cent better a day, which compounds over time. That suddenly makes the size of your opening ambition unimportant and repeatability decisive. Hence the two-minute rule further on: better a habit so small that it can hardly fail than an ambition that holds out for three days.

And the third reason is the reason most people drop out: results lag behind behaviour. The plateau briefing further on, in section 03, is about precisely that. Without that warning up front you read the silence in week four as proof that it is not working, while the proof is simply not visible yet.

For anyone who wants to apply this line of thinking more widely than habits: the SMART goals coach sits on exactly the other side of this debate and sharpens goals instead of replacing them. Using both is fine, as long as you know which tool you are holding. And if the habit you are building is learning something properly, the AI Feynman Technique Coach is the method that tests whether you really understand it.

A skill is no more than its instructions, so we simply describe them here. The file opens with frontmatter: name, version 1.0.0, MIT licence and a long description that decides when Claude picks the skill up. Then comes the role: a coach who applies the framework of James Clear from Atomic Habits (Avery, 2018), holding on to three convictions. Goals lose to systems, identity comes before behaviour, and 1 per cent better a day delivers exponential gains over time.

The core consists of four ideas. Identity-based habits, the habit loop with four links, the four laws of behaviour change in two directions, and the plateau of latent potential. Those four do not just pass by as theory, they decide the structure of the answer you get.

The method counts six steps, in order, with no skipping. Step 1 establishes the identity: who do you want to be in essence, and that gets written down literally in the output. Step 2 is the diagnosis of your existing loop, written out as a table with four columns. Step 3 turns the four laws into concrete actions, and the file is strict about that: no vague advice, but workable behaviour for tomorrow morning.

Step 4 is habit stacking with the formula after this existing habit I will do that new habit, because an existing habit already runs automatically and can therefore serve as a cue. Step 5 is the plateau briefing, with the days attached. Step 6 agrees a weekly review with three questions: what worked, what did not, what will you do differently this week.

The output format is fixed. Inside it sits a complete template with fixed headings in a fixed order: Identity, Loop diagnosis as a table, The Four Laws applied with a subheading per law, Habit stack, Plateau warning and Weekly review with three lines to fill in. That way you get the same kind of document for every question, and not a list one time and an essay the next. Under that template sits the full worked example that you see shortened in FIG.01 above.

Then comes the part that most decides how the coach sounds. Be direct and concrete, avoid vague motivation. Address the user on their identity, not on their willpower. Never give plans longer than two weeks. Point to the book when someone wants to go deeper, because a book summary in the chat is not the intention. And on relapse: never shame, a relapse is data. Ask what the cue was, how the surroundings should have looked and what friction was at play, and only then adjust the system, not the goal.

Right at the end it names four sources it leans on. Deliberately so: you can read them and decide for yourself whether you agree with them. They are set out below in DOC.04.

The source list in the file counts four titles, and together they explain why the skill does what it does. They are not there for show: the file says explicitly that this is not a magic potion but an accessible synthesis of existing behavioural research.

James Clear, Atomic Habits (Avery, 2018), is the main source. From that book come the four laws, the idea that identity comes before behaviour, habit stacking, the two-minute rule and the plateau of latent potential. Clear did not write a scientific work but a practical framework, and it is that practical form which makes him suitable to pour into an instruction file: every part is an action.

Charles Duhigg, The Power of Habit (Random House, 2012), supplies the loop. Duhigg popularised the thought that behaviour sits in a recognisable cycle you can take apart and rebuild. The diagnosis table in step 2 of the skill is directly derived from it: first make the loop visible, only then intervene.

BJ Fogg, Tiny Habits (Houghton Mifflin, 2019), supplies the small step. Fogg researched at Stanford how behaviour comes about and arrived at shrinking the action instead of raising the motivation. Which is precisely what the two-minute rule does. To use the model behind that thinking on its own, there is a separate Fogg behaviour model coach skill for it, which splits behaviour into motivation, ability and prompt.

B.F. Skinner supplies the foundation. His classic work on operant conditioning explains why the fourth link, the reward, closes the loop and strengthens the behaviour. It also explains why the skill is so strict at law 4 about which reward you choose: a reward that contradicts your identity strengthens the wrong loop.

What is deliberately absent is at least as relevant. Nowhere does it claim success rates, no number of days after which a habit would supposedly become automatic, and no research figures about who keeps it up and who does not. We do not adopt any either. What you get is a framework with a provenance, not a promise with a number.

The zip is 4.2 KB and contains one folder, ai-atomic-habits-coach, with the file SKILL.md inside it. Nothing else: no README, no scripts, no attachments. Installing is therefore no more than putting that folder in the right place, and that place differs per environment. SKILL.md has been an open standard since December 2025, so the same file also works in Codex, Cursor and Gemini CLI. You are not downloading a Claude file but a working instruction that any modern AI assistant can read.

After that, use is simple: describe the habit you want to build or want rid of, in plain language. The more concrete your situation, the more usable the diagnosis, because the skill can only take apart a loop whose parts you name. If you do not know where to start, simply name the moment: where you are, what time it is and what you did before. It asks for the rest itself.

If you get stuck on the installation itself, the full explanation per environment is in installing Claude skills. And if you want to understand why this kind of file works the way it does, what Claude skills are is the best starting point. The broader articles about working with AI are in the knowledge base.

The skill is at its strongest when you have a concrete behaviour in mind and it just will not stick. Reading every day, getting up earlier, scrolling less, exercising weekly, a fixed block for concentrated work. Right there the whole machinery comes into its own: an identity that fits, a loop you recognise, four laws you can carry out tomorrow morning and a stack that hooks onto something that already runs by itself.

It works just as well the other way round, on habits you want rid of. It then inverts the laws and looks not for more discipline but for less opportunity. Note how far that sits from the average advice: the answer to too much scrolling is rarely more willpower and usually a phone lying in the hallway at a quarter to nine.

There are also situations where you are better off leaving it alone. The first is named by the frontmatter itself: do not use it for loose motivational quotes or one-off coaching. Anyone looking for a lift gets a system here, and that feels barer than expected. The second limit is heavier. If there is a problem underneath the habit such as depression, addiction or an eating disorder, the skill refuses to build a habit plan that tidies the symptom away, and refers you to professional help. That is a limit the file draws itself and one we do not soften here.

A third limit is practical: it does not plan more than two weeks ahead. Want a plan for the whole year? Then a habit coach is the wrong tool and what you need is a goal framework instead. And the last honest limit: the skill does not change your behaviour, it designs the system in which that behaviour becomes easier. Showing up stays your job. What it does do is stop you giving up at the moment when the system was in fact already working.

If the subject fits focus better than habits, look at the Getting Things Done coach or the Eisenhower matrix coach. If it is about intentions and obstacles in advance, the WOOP method coach is the logical neighbour.

What holds for a person holds just as well for a company: what happens structurally is almost always a system and almost never motivation. Publishing every week, asking for a review every month, every quote out of the door within two days. Do that on willpower and you keep it up for a month. Anyone who makes a system of it does it for years.

For your own habits the weak spot sits somewhere else: Claude starts every session empty, so you supply your identity sentence, your four laws and your previous plateau again every time. That is exactly what Jarvis solves, the organisation brain that starts at €9 per month including Dutch VAT. Should your team learn to set up this kind of method themselves instead of one person doing it, that is the work of the AI training.