The AI Intake Summary skill for Claude
You paste three pages of scribble, a voice note or a full transcript into Claude and you get back an intake report somebody else can work with straight away. Conclusion at the top, the facts separated from your interpretations, the client's quotations kept word for word, and at the bottom the exact questions you still have to ask. What the skill does not know it does not write smooth, it marks it as unknown. That is not a limitation, that is the whole point.
the lines above come from the zip on this page · SKILL.md is 13,867 bytes
Scribble in, a report out
reg. I.001This is the worked example that sits in the SKILL.md itself, shortened here. The input is a note as you really make one after a call: lower case, half sentences, a quotation from memory. Watch what happens to the budget. Robin only named an upper limit, so the field stays on unknown and the skill tells you why.
What the AI Intake Summary is
The AI Intake Summary is a free skill from the library of 100 skills of TheSEO. A skill is an instruction file, SKILL.md, that gives an AI assistant a fixed method for one task. What a skill is exactly and why that format works so well you can read in what Claude skills are; this one is a text file of 2,342 words that tells Claude how to turn raw conversation input into a report somebody else can carry straight on with.
No software, no subscription, no link to your calendar: you download the zip at the top of this page, put it in your Claude environment and from that moment on Claude writes your conversations up according to this structure.
The problem the skill solves you probably know from your own week. You come out of a good conversation, you have three pages of keywords and a head full of context, and then the writing up begins. An hour later you have a report that recounts chronologically what was said, in which your own assumptions have quietly been written down as facts, and about which your colleague asks next week what was actually agreed. Or you do not write it up, and two weeks later you are standing over the quote wondering whether he did or did not mention that budget. Exactly those sighs are in the file literally as moments when the skill should kick in.
The instruction the file gives Claude is strict and simple at once. It opens with an identity of three sentences: you are an English-language intake analyst, you add nothing that was not there, and you make what is missing visible instead of writing it smooth. Everything that comes after, from the theory to the output blocks, is a working out of those three sentences.
This one is written for anyone who has client conversations and then has to deliver something: sole traders and freelancers, agencies, consultants, building and installation firms, coaches. Along with the rest of the skill library we make it available for free from the AI and automation service, with no account and no sales email afterwards.
Why conversation notes lose their value
Notes do not lose their value because you write badly, but through three creeping mistakes that almost everyone makes. All three are built into the method here, so it pays to have them sharp.
Mistake one: the report is chronological. You write down what happened, in the order in which it happened. It feels complete, but nobody reads it to the end, and the most important conclusion sits wherever the conversation happened to end. The file is hard about it: an intake report that recounts a conversation chronologically is unusable. That is why the skill enforces the pyramid structure, with the conclusion in the first five lines.
Mistake two: assumptions disguise themselves as facts. Between the conversation and the writing up sits your own head, and it fills gaps effortlessly. The client said quotes take three days; you write down that he loses revenue because of it. Perhaps true, but he did not say it. The moment that sentence is in the report, everyone treats it as a statement from the client, and you build a proposal on a foundation that is not there. Those two streams stay strictly separated here, and every inference is marked as an inference.
Mistake three: vague stays vague. Soon, shortly, reasonable budget, big problem: words that felt clear in the conversation and mean nothing on paper. Each one gets translated into something measurable or marked explicitly as unknown. Soon becomes within two weeks or unknown, never quietly filled in.
In diff form, with lines pointing to the example from the skill: what a smoothed-over report says, and what the skill makes of it instead.
Ten blocks, a fixed order
reg. I.002Every summary the skill delivers has the same ten blocks, in the same order. Anyone who reads only the first block must be able to retell the conversation; every following block adds depth for anyone reading on. The mono lines underneath say why the block sits there.
What is really in the SKILL.md
A skill is only as good as its instructions, so we simply describe them here. The file opens with frontmatter that says when Claude should pick the skill up. Not only at direct requests such as summarise an intake, write up a call report or work up my notes, but also at sighs: I have three pages of scribble and no idea any more what was agreed, I lose an hour writing up after every conversation, my colleague cannot follow my notes, every report looks different. Say something like that to Claude while the skill is loaded and you get it automatically.
Then comes the heart: seven fixed actions the skill carries out for every summary. It reads the raw input all the way through and marks what is fact, what is a quotation and what you have already interpreted yourself. It sorts everything into the four SPIN layers and leaves a layer empty if there is no information for it. It puts the conclusion at the top, according to the pyramid. It translates vague indications into something measurable or marks them explicitly as unknown. It delivers a phased approach in outline. It delivers price framing without inventing amounts. And it closes with the list of missing information, including the exact questions to go and get it.
Before anything is written, the file holds a short input checklist of six points: the raw notes themselves, who the client is, what you sell or deliver, what the summary is for, whether things were agreed that are not in the notes, and whether a deadline or a follow-up appointment was mentioned. The instruction attached is typical: only ask about what is missing and keep it short, because this skill is often used by someone who has just come out of a conversation. Paste only notes with no context and it simply carries on, putting its assumptions visibly in a separate block.
A striking number of rules are about language. Clear English addressing the reader as you, short sentences alternating with long ones, a figure above an adjective, and quotations from the client in quotation marks and word for word. Do not smooth them out, says the file, because the client's own wording is the most usable material for the quote. There is also a separate review mode: give it an existing call report or intake form and you get a completeness test per SPIN layer, a marking of every assumption written down as a fact, a vagueness scan, the qualification gaps at the top and a rewritten version according to the standard structure.
The file closes with a source list of five entries, from Rackham to Weiss. More on that below, because that theory is the reason the report is built the way it is built. If you want to learn to write such an instruction file yourself, the full method is in writing a SKILL.md.
Four layers, one weak spot
reg. I.003Every conversation gets sorted into the four information layers of SPIN, with the four qualification points of BANT tracked alongside. The figure is filled in with the example of Robin from the SKILL.md. Look above all at the third layer: that is, according to the file, the part that later decides whether a proposal is accepted, and at the same time the worst recorded part of almost every conversation.
# filled in with the worked example from the SKILL.md · the source comes with every point: said, inferred or unknown
The theory the skill rests on
According to its own explanation the skill combines two recognised frameworks: one for what you take out of a conversation, and one for how you write it down. The SKILL.md names its sources, and that is deliberate: it lets you read them and decide for yourself whether you agree.
The content comes from Neil Rackham. For what has to be taken out of a conversation the skill uses the logic of SPIN, described in Spin Selling in 1988. Rackham analysed large numbers of sales conversations and found that successful conversations produce four kinds of information: the facts, the problem in the words of the client, what that problem costs, and what it delivers once it is solved. Holding the conversation itself according to that method is a different craft from writing the report, and there is a separate skill for it: the SPIN Sales Conversation skill.
The form comes from Barbara Minto. Her Pyramid Principle, developed at McKinsey and described in 1987, says: start with the conclusion, then the supporting points, and only then the details. Hence the heart of it in five lines at the top of every report. A reader who reads only that block can retell the conversation; everything else is depth for anyone who needs it.
The qualification comes from BANT, and the file is honest about it. Budget, Authority, Need, Timing is generally attributed to IBM and has no traceable original publication. Not a model but a checklist, says the SKILL.md literally, and that is exactly how it uses it: four points, per point known or unknown, with the source attached. For complex B2B processes with several decision makers the file names MEDDIC as a heavier alternative, with the note that most small business intakes do not need it.
The handover comes from SBAR, the price question from Weiss. SBAR, from the United States Navy and widely applied in healthcare since the start of this century, is the compact variant for an internal handover: Situation, Background, Assessment, Recommendation in about half a page. And from Alan Weiss, author of Million Dollar Consulting, comes the thinking behind the price framing: ask about the cost of the problem before you name a price. Anyone who wants to understand what the client is really trying to achieve behind the question they asked can lay the report beside the Jobs to be Done Analyst skill .
You can improve your own reports with these principles without installing the skill: put the conclusion at the top, separate what the client said from what you assume, and write down what you do not know instead of writing around it.
What the skill refuses
reg. I.004The SKILL.md holds a list of seven things the skill never does, and that list matters at least as much as what it does do. An intake report is the source your quote, your planning and your handover build on: if something is in it that was not said, you build everything on quicksand. In conversation the rules play out like this: every rule is a request you could make, with the response the skill gives according to its own instructions.
Installing in Claude Code, Claude.ai or Codex
The zip contains one folder with the SKILL.md inside it. Installing is a matter of putting the file in the right place, and that place differs per environment. SKILL.md has been an open standard since December 2025, so the same skill also works in Codex, Cursor and Gemini CLI. You are therefore not downloading a Claude file but a working instruction that any modern AI assistant can read. The full guide per environment, with screenshots, is in installing Claude skills.
- Unzip it into
~/.claude/skills/(or.claude/skills/in your project). - Claude then recognises the skill by itself as soon as you start on an intake or a call report.
- You can also call it directly, with
/ai-intake-summary.
- Go to Customize and then Skills.
- Upload the zip there as a skill.
- Or paste the contents of SKILL.md into the project instructions of a Project.
- Open
AGENTS.mdin your repo. - Paste the contents of SKILL.md into it, or put SKILL.md beside it as a separate file and point to it from
AGENTS.md. - Codex reads that along at every session.
After that, use is simple: paste your notes straight after the conversation and ask for a summary. The rawer the input, the more the structure delivers; tidying up first is a waste of your time. Whatever is missing, it asks about itself, and keeps that short. New to this kind of file? The knowledge base holds the broader explanation of working with AI.
When you use it and when you do not
According to the file the skill has several modes, and picks the form to fit the conversation. You use the full structure with all ten blocks for an intake leading to a quote. For an internal handover to a colleague or project lead it switches to the SBAR variant of about half a page. After a short phone call the shortened variant of three blocks is enough: the heart of it, what is missing, the next step, because nobody reads a two-page report about a seven-minute conversation.
For processes with more than two decision makers or a tender a decision maker analysis is added, with what each person wins or loses from a yes. Work with a word-for-word transcript and the quotations get time stamps. And with repeat conversations with the same client the report starts with what has changed since last time.
There are limits too, and they are in the file itself. Writing a full quote is not part of it: that is a separate step after the summary, with its own document and its own logic, and there is the AI Quote Writer skill for it, which builds on exactly this report. Files where legal requirements apply to record keeping, medical or legal, are out of scope as well: there the law decides the format, not the readability. And it does not work with recordings for which no consent was given.
The most honest limit is this: your report gets better, your conversation does not. Ask the wrong questions and the report shows that mercilessly, because the empty layers and open fields are visibly in it. That is a hard swallow, but it is exactly the feedback that makes your next conversation better. When the price comes up later, the Negotiator skill helps with the conversation that follows the proposal.
Run it yourself or have it run
reg. I.005This skill is the free do-it-yourself version of work we also deliver as a service. It stays complete and free of catches, but understand what a skill is: it teaches your AI how to do something, while every new session starts empty. A skill is not the engine and not the memory. You prompt, you paste your notes and your context again every time, you check. Anyone who wants that differently has two next steps: hand over the engine, or sort out the memory.
where you are now The skill: you are the engine You run the AI Intake Summary yourself in Claude, Codex or Cursor. It costs nothing, it works today, and you keep it entirely in your own hands: no trial period, no locked-off parts. The limit is your own time: it only happens when you prompt.
have it prepared The AI employee: it is ready without you prompting If your bottleneck is not one report but the whole of the sales groundwork, there is the AI employee as a service. For sales work Mansotti, the company of which TheSEO is the trading name, builds the Sales employee, who does prospect research; alongside that there are the Quote employee (preparing enquiries and quotes) and the Reporting employee (weekly and monthly reports). The control stays with you, because output remains a draft until a human approves it.
everything from one source Jarvis: all your AIs work from the same company knowledge The skill teaches the AI, the brain is where the memory lives. If you want all your AIs to work from the same company knowledge, that is Jarvis, the organisation brain. It connects ChatGPT, Claude, Codex and your people to the same projects, core knowledge and decisions, so your next AI session does not start over. For this skill that means: you do not have to explain again per session who the client is and what was discussed last time, because earlier reports and agreements are already there. What that delivers in practice, from the plans to your first week, you can read at Jarvis itself.
What Jarvis actually delivers
reg. I.006Step 3 deserves more than a paragraph, because with intakes in particular memory is the difference. Making a report is one thing; finding it again the moment the client calls three weeks later is where it gains its value. Jarvis is the organisation brain: it remembers what your AIs need to know, divides the work and keeps track of what happened. You notice it first at the follow-up conversation.
We have been running on this system ourselves for months. Every agent session, every task and every decision is logged in it and can be read back. A new session therefore does not start blank: it first fetches the recorded decisions, the running projects and the latest changes, and carries on where the last one stopped. For intakes that is the difference between eight separate reports and one running record in which the open question from three weeks ago is still open. So we are not describing a promise but the way we work every day.
All plans are at jarvis/pricing. You pick your own plan. Paying online opens as soon as the last checks are done; until then we set your account up on request. For business agreements and bespoke work there is a conversation.
The skills around it
reg. I.007An intake report sits in the middle of a chain: a conversation comes before it and a proposal comes after it. These skills from the same library of 100 each take a different piece of that line.
Before and during the conversation
the inputThe report is only as good as the conversation. These two make sure there is something to note down.
SPIN Sales ConversationThe questioning technique that gets the cost of the problem on the table: exactly the layer missing from most notes.SKILL Jobs to be Done AnalystLays bare what the client is really trying to achieve behind the question they ask.SKILLAfter the report
what followsThe summary is the foundation. These two build on it.
AI Quote WriterTurns the intake report into a quote with three price options and a next step with a date.SKILL NegotiatorFor the conversation that follows the proposal, when the client wants to talk about the price.SKILLThe foundations
understand it firstNew to skills? These three articles explain the format.
What Claude skills areWhat a SKILL.md is, why the format works and what you can do with it.DOC installing Claude skillsThe full installation guide per environment, step by step.DOC writing a SKILL.mdBuilding a skill yourself according to the open standard.DOCLooking further
the contextWhere this skill comes from and what else there is.
The whole skill libraryAll 100 free skills in a row, sorted by subject.HUB AI and automationThe service behind it: from separate skills to working automation in your business.SRV AI trainingIf your team wants to learn to set up this kind of work themselves.SRV Knowledge baseArticles about SEO, AI and online visibility, searchable.DOCFrequently asked questions
What does the AI Intake Summary skill cost?
Nothing. The download is free, falls under the MIT licence, and you do not have to create an account or leave an email address. You download a 5.8 KB zip containing one folder and a single file, SKILL.md, and that is the complete skill. There is no paid version and no sales email follows.
Does this skill work in Codex, Cursor or Gemini CLI?
Yes. SKILL.md has been an open standard since December 2025, so the same file works in Codex, Cursor, Gemini CLI and other tools that follow the standard. In Codex you unzip it into .agents/skills/ in your project, or into ~/.agents/skills/ for all your projects; Codex has supported SKILL.md directly since the open standard of December 2025. Pasting the contents of SKILL.md into your AGENTS.md still works too. The instructions are plain readable text, so any assistant that accepts instruction files can handle it.
Does the skill invent information that is not in my notes?
No, and that is the first rule in the file: the skill adds nothing that was not there. If the budget is missing, it says unknown. If a consequence is your own estimate, it is marked as inferred and not as a statement from the client. What is missing is made visible instead of written smooth, because a report that looks more complete than the conversation was is more dangerous than an honest report with open fields.
Can I use a voice note or a full transcript as well?
Yes. According to its own description the skill works with keywords, half sentences, voice notes and full transcripts. With a word-for-word transcript it puts time stamps on the quotations, so you can listen back to the exact moment later. The rawer the input, the more important the structure that goes over it, so you do not have to tidy anything up before you paste it.
Can the skill assess an existing call report?
Yes. Give it an existing report or intake form and you get an assessment in five steps: a completeness test per SPIN layer, a marking of every sentence where an assumption is presented as a fact, a vagueness scan on words such as soon and reasonable budget, the qualification gaps at the top instead of as a footnote, and a rewritten version according to the standard structure with the missing points explicitly marked as unknown.
Does the skill name prices or set my rate?
No. The price framing is deliberately a framing without amounts: the skill sets out which range the client named themselves, which point of comparison they brought up, where the price sensitivity sits and which anchor is logical. The amount itself you decide. Inventing figures, amounts and rates is literally in the list of things the skill never does.
Which conversations is the skill not meant for?
For three situations, and they are in the file itself. It does not write a full quote, because that is a separate step after the summary. Nor is it meant for medical, legal or other files where legal requirements apply to record keeping. And it does not work with recordings for which no consent was given.
First conversations, then reports
A better intake report makes everything that follows stronger, but there do have to be conversations to summarise. When too few enquiries come in, the real work sits before this document: being found by the people looking for you. The free SEO scan shows in a few seconds where your site stands. And if you want to talk further about what AI can do for your business, from separate skills to full automation, we simply do that in a conversation.