EN
Language · same page NLNederlands/ai-en-automatisatie/ai-acquisitiemail-schrijver-skill-voor-claude/ ENEnglish (UK)/en/ai-and-automation/ai-acquisition-email-writer-skill-for-claude/ ESEspañol/es/redactor-de-correos-de-prospeccion-habilidad-para-claude/ We do not remember your choice and never redirect you automatically.
~/skills/ai-acquisition-email-writer[ok] loaded
INS.SKILL · FREE FOR CLAUDE, AND FOR CODEX AND CURSOR TOO

The Acquisition Email Writer skill for Claude

Writing an acquisition email starts with what you know about the recipient, not with what you are selling. This skill asks about that first, tests your observation against three requirements, and only then writes. You get a subject line that covers what is in the email, an opening about their company, a core of four sentences, and a closing question that can be answered in one word. If you do not know anything concrete, it refuses to write and asks you questions instead. No invented mutual acquaintances, no Re: in the subject line, no deadline that does not exist.

Download the skill View all 100 skills NL 9.3 KB · zip with SKILL.md · MIT licence · no account needed
$ installation, line by lineFIG.00
~/.claude/skills
theseo ~ %
  inflating: ai-acquisitiemail-schrijver/SKILL.md
theseo ~ %
  ---
  name: ai-acquisitiemail-schrijver
  description: 'Schrijft een koude acquisitiemail in het Nederlands die begint…
theseo ~ %
[ok] skill active · 4,026 words of instructions loaded

the lines above come from the zip on this page · SKILL.md is 24,910 bytes

~/skills/ai-acquisition-email-writer/01-demonstration[run]
Section 01 · The demonstration

One observation in, one email out

reg. A.001

This is the first worked example from the SKILL.md itself, shown here in shortened form. Notice what does not happen: no second trigger is invented to pad out the email, no percentage is added that nobody mentioned, and the closing question stays singular. What the user has not said appears as [FILL IN].

FIG.01: From observation to a ready to send email[#----] sheet 1/5
~/acquisition · claude + ai-acquisition-email-writer
you > write an acquisition email to this company
input.txtthe input from the example I do interim production planning for manufacturing companies. Bracken Windows, around 40 staff. Their own vacancy page has had a production planner role open for three weeks, and the text literally says that delivery reliability is under pressure. I previously spent six months at a window and door company. I do not know the contact person. Address from their website.
claude · testing observation, building email, checking itself...
[1/11]observation · supplied: vacancy text with a sentence about delivery reliability, source included
[2/11]test · specific: yes · traceable: yes, their own vacancy page · useful: yes, touches your work
[3/11]evidence · six months at a window and door company; no outcome supplied, so no figure
[4/11]recipient · unknown; assumption at 40 staff is the director, noted in the notes
[5/11]subject · "Your production planner vacancy" · four words, covers the content
[6/11]opening · starts with their vacancy page; the word I does not appear in the first sentence
[7/11]core · three sentences, with the acknowledgement that you do not know their process
[8/11]closing question · "Has this already been assigned to someone at your end?" · answerable with yes or no
[9/11]sender block · name, company, KVK, phone number and an opt-out line in plain language
[10/11]self-check · 11 tests run, for this example 5 corrected and 6 passed
[11/11]delivery · email, check block, notes and one follow-up email with a different angle
email.txtshortened, the full output is in the skill Subject: Your production planner vacancy Hi [FILL IN first name], Your vacancy page has had a production planner role open for a few weeks, and the text says that delivery reliability is under pressure. That is an honest line to put in a vacancy, and it usually means your planning has been stretched too thin for a while. I do interim production planning at manufacturing companies, often in exactly that gap between a planner who has left and a new one. At a window and door company I spent six months doing that. I do not know whether your situation is similar, because I do not know your process. Has this already been assigned to someone at your end? Note: your evidence still rests on "I spent six months doing that there". A sentence with an outcome you are allowed to name would make this email stronger
[######] 11/11 done · 102 words, one question, nothing invented
11A.01 · fixed steps per email
11A.02 · tests in the self-check
12A.03 · things the skill refuses
0A.04 · euros, MIT licence, no account
A.01 to A.04 · properties of the skill file, not results or usage figures.
$ cat 01-wat-de-skill-is.md
DOC.01

What the Acquisition Email Writer skill is

The AI Acquisition Email Writer is a free skill from our skill library, which now counts 100 skills. A skill is an instruction file, SKILL.md, that gives an AI assistant a fixed way of working for one task.

No software to install, no subscription, no login: a single text file of 3,972 words that tells Claude how to build a cold business email, what it needs to know from you first, how it checks its own draft before delivering it, and what it never writes under any circumstances. You download the zip at the top of this page, put it in your Claude environment, and from that moment Claude works to this structure. What a skill actually is and how that file format works, you can read in what Claude skills are NL.

You recognise the problem this skill solves instantly if you have ever written or received an acquisition email. The writer starts with themselves. My name is so and so, we are an agency in such and such, we help companies like yours with this and that. Only on line four does something about the recipient appear, and that is usually a compliment that fits any company. The recipient knows within two seconds that this is an email that also went to a hundred others, and then it is over. Not because the service is not good, but because there is nothing in the email that is only about them.

Here it works exactly the other way round, with one hard rule: the observation first, only then the email. An observation is something concrete that you have seen or read somewhere about this one company. A vacancy that has been open for weeks. A sentence on their own site. A step in their ordering process where you yourself got stuck. A new branch. A page that is broken. That observation is then tested against three requirements: specific, so it must not fit just as well on any other company in the sector; traceable, so you can say where you saw it; and useful, so there is a conceivable link with what you do. If the observation fails, it writes nothing and you get questions back instead.

That immediately makes it unsuitable for a certain kind of work, and that is deliberate. If you have a list of two hundred addresses and want to pour one text over all of them, this is not your tool. It will not start on that, it advises against it, and it does not help you compile such a list either. What it does do is write ten emails that do not resemble each other, provided you answer three questions per company. Expect it to take work. It is just the work that makes the difference between an email that gets ignored and an email that gets answered.

It is useful for anyone who has to bring in clients themselves without a sales department: freelancers, small agencies, consultants, installers looking for business clients, software builders, and anyone who comes home from a trade show with a stack of business cards. This skill belongs to the skill library NL we make freely available through our AI and automation service, with no account and no sales email afterwards. That last part is the least you can expect from a skill about acquisition emails.

One thing sits right at the front of the file, and we quote it here word for word: this skill is not a lawyer. It knows the broad outline of the Dutch rules around unsolicited business email, it puts the opt-out line and the recognisable sender block into your email as standard, and it refuses a misleading subject line. But it never judges whether the way you collect addresses and send email is lawful. For that you need your own lawyer, and the file points you there too.

$ cat 02-waarom-het-misgaat.md
DOC.02

Why acquisition emails go unanswered

There is a separate assessment function built in for emails you have already written, and its checkpoints read like a list of what goes wrong in practice. Each one is recognisable on its own, and together they explain why an email that reads perfectly well on your screen does nothing in someone else's inbox.

The email starts with the sender. According to the file, this is the most common problem, and it is also the easiest to fix. The first sentence is about you, your company or your offer, while at that point the reader has zero reason to want to know anything about you. The file solves this with a test that leaves no room: if the email starts with I, we, my, our or your own company name, the first sentence gets rewritten.

The email survives the name swap. Replace the name of the receiving company with any other and read the email again. Does it still hold up? Then you have not written a personal email but a template with a name dropped in. In the file this is called the copy test, and it is the only test where the skill does not correct but stops completely. There is simply nothing to rewrite: there is just no trigger yet.

The offer comes before the trigger. Many emails neatly explain what the sender does and for whom, but never make the connection to this specific company. The reader has to build that bridge themselves, and they do not. This is what the file devotes the core of the email to: a maximum of four sentences that tie the observation and your work together, including one mandatory sentence that acknowledges what you do not yet know. That last part sounds weak and is the opposite: it is the one small sentence in the email that proves you are not bluffing.

The question is too big. Shall we grab a coffee sometime to see what I can do for you. That is not a question but a calendar request, and answering it costs the recipient half an hour of thinking. Every email therefore ends with one question that can be answered in one word, and where no is a perfectly fine answer. Someone who answers no has answered you, and that is more than you get from most emails.

There are tricks in it. Re: in the subject line to suggest a conversation that never happened. A deadline that does not exist. A mutual acquaintance you do not actually know. Only two spots left this month. It sometimes works once, and never again with that person after that. The skill categorically refuses these tricks, even if you explicitly ask for them, and it states why in the check block.

In diff form, with the difference the assessment function marks in almost every email it is given: what comes out and what goes in instead.

~/skills/ai-acquisition-email-writer/02-anatomy[ok]
Section 02 · The anatomy

Six blocks, each with its own rule

reg. A.002

An email from this skill consists of six blocks and every block has one rule that is never departed from. On the left is the block, in the middle the text from the second example in the SKILL.md, on the right the rule from the file that governs that block. What is yellow is a field the skill does not fill in for you because it is your own detail.

FIG.02: The anatomy of the email, block by block[##---] sheet 2/5
anatomy.mdexample 2 from the SKILL.md · online shop without guest checkout
BLOCK 01subject Setting up an account to check out three to six words, literally about the content. No Re:, no Fwd:, no exclamation mark. The recipient must not feel misled when they open it
BLOCK 02opening Hi Mark,
Yesterday I tried to place an order with Bruning Bike Parts and got stuck at the point where you have to create an account before you can pay.
the first sentence is about the recipient and contains the observation plus where you saw it. The word I may only appear here if it is about your own action on their site
BLOCK 03core I work on ordering processes for online shops. At an online shop selling bike lights, I removed that mandatory step and after that the share of abandoned orders went down. I do not know if it would work out the same way for you, because I do not know how many visitors you have or why you built that step in. a maximum of four sentences, with one piece of evidence the user has actually supplied and one sentence that acknowledges what you do not yet know. No amounts in a first email
BLOCK 04closing question Was that a deliberate choice? exactly one question, answerable in one word, and no must be an acceptable answer. Two questions is one question too many
BLOCK 05sender block Regards,
[FILL IN name], [FILL IN company]
KVK [FILL IN], [FILL IN phone number]
the sender must be recognisable: real name, real company, KVK number. The skill never fills in your details itself and never invents numbers
BLOCK 06opt-out line If you would rather not get email about this, just reply no. Then you will not hear from me again. a working opt-out option in plain language, and you stick to it. This block never gets dropped, not even if you ask for that
# below that come three more blocks that do not go to the recipient: CHECK with the outcome of the eleven tests, NOTES with your assumptions and open fields, and FOLLOW-UP EMAIL 1 with a different angle
$ less ai-acquisitiemail-schrijver/SKILL.md # 3,972 words of instructions
DOC.03

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 states when Claude should pick up the skill. Not only for the obvious terms like write acquisition email, cold email, prospect email or write a follow-up email, but also for sentences people actually use: I never get a reply to my emails, it feels pushy to email someone, my email sounds like a flyer, I have a hundred addresses and no idea what to write, what do I put in the subject line, how many times am I allowed to remind someone.

Anyone who says something like that to Claude while the skill is loaded gets it automatically. The same frontmatter also lists the cases where the skill should specifically not start, and that is a deliberate choice: over-triggering is the most common complaint about popular skills. Below that sits the licence: MIT, version 1.0.0.

Then comes the core: eleven fixed steps. It starts with the question of what you concretely know about this company and where you saw it. It tests that observation against specific, traceable and useful. It asks what you deliver, for whom that worked before and what changed then. It asks who the recipient is and whether that person is the right one to decide on this. Then it builds the subject line, the opening, the core and the closing question, in that order and each with its own rule. It puts the sender block in place with the opt-out line underneath. It runs the self-check. And it delivers the email, the check block, the notes and one follow-up email.

Before anything gets written, the skill runs through a mandatory eight-point input checklist: the observation with where you saw it, the company name and size, the recipient and their role, what you deliver and what not, the evidence you actually have, where the address came from, whether there was earlier contact, and the tone.

That point about where the address came from is sharper than it looks. It goes into the email if the recipient could reasonably ask about it, and at the same time it is your own test: if you cannot answer that question, you already know enough. If the observation is missing, the skill stops. If something else that is not crucial is missing, it carries on and puts its assumption visibly in the notes.

There is a separate chapter about language, and it is written for Dutch and not translated from English. An email stays under 150 words and preferably under 120. Short sentences alternate with long ones. No dashes in running text. And there is a list of openings that are completely banned, which in English come down to: I am reaching out because, my name is X and I work at Y, we are a leading agency, I noticed you have a great company, hope this email finds you well, I will keep this short.

Anyone who cuts those six sentences out of their own emails is already well on the way, even without this skill. On top of that, the file strips out the well-known AI tells, which is exactly the same work the Text Humanizer skill does, just built into the writing itself.

There is also an explicit rule about prices in it: they do not belong in a first cold email. If you ask for one anyway, the skill puts the amount in with excluding VAT next to it, and notes in the notes section that it advises against this and why. Here the skill does what you ask while also having an opinion about it, which is rare in this file. The proposal with the prices in it comes later, and that is what the Offer Writer skill is for.

The file also has a second mode. If you hand it an email you have already written, the skill assesses it on five points: first the copy test, then the I-you count with both numbers next to it, then a marker on every sentence the recipient does not need in order to reply, then a verdict on the subject line and the closing question with an alternative, and finally a rewritten version plus the questions you still need to answer before the email is genuinely personal. If you want to learn to set up an instruction file like this yourself, the method is at writing a SKILL.md NL.

~/skills/ai-acquisition-email-writer/03-self-check[ok]
Section 03 · The self-check

The skill attacks its own draft

reg. A.003

This is the part that sets this skill apart from a prompt that simply asks for an email. A first draft is never delivered. It writes first, then attacks its own text with eleven tests, corrects whatever fails, and lists under your email which test failed and what was changed. Below is that log as it came out for the second example in the SKILL.md, including the tests that flagged something there.

FIG.03: check.log · eleven tests on the draft itself[###--] sheet 3/5
check.logresult for example 2 · online shop without guest checkout
01first-word testDraft started with "I do conversion optimisation". Rewritten so the first sentence starts with their ordering process.CORRECTED
02copy testCompany name replaced with any other: the opening no longer holds up. That is good news.PASSED
03invented-claim testDraft contained "you are leaving revenue on the table here" with a percentage after it. That figure had not been supplied and has been removed.CORRECTED
04I-you countThe draft had six I-forms against two you-forms. Core rewritten, now five against six.CORRECTED
05subject-line testThe requested Re: was not used. The subject line now names what the email is actually about.CORRECTED
06question testOne question, answerable with yes or no, and no is a perfectly fine answer here.PASSED
07urgency testNo deadline, no scarcity, no urgency in the text.PASSED
08cliche testBanned openings and AI tells checked, nothing found in this draft.PASSED
09length test106 words. Under the limit of 150, so nothing needs to be cut.PASSED
10legal testKVK line and opt-out line were missing from the draft and have been added. No verdict on lawfulness.CORRECTED
11read-aloud test"Some of your visitors are guaranteed to drop off there" became "often", because you cannot know it is guaranteed.CORRECTED
[###########] 11/11 run · 6 corrected, 5 passed · if the copy test or the invented-claim test fails, the skill delivers questions instead of an email
$ cat 04-bronnen-en-wet.md # Minto, Rackham, Voss, Cialdini, ACM
DOC.04

The theory and the legal foundation

The choices in the file can be traced back to work you can look up yourself, and where something is craft rather than research, the file says so. The structure of the email follows Barbara Minto's Pyramid Principle from 1987: the trigger up front, the explanation behind it. Anyone who spends their first paragraph on themselves has lost the reader before the reason to keep reading appears.

The question order comes from SPIN Selling by Neil Rackham, 1988. Rackham describes how, in a sales conversation, you first ask about the situation and the problem before talking about a solution. A cold email is too short for that whole sequence, so the skill uses part of it: the observation plays the role of the situation question and the closing question plays that of the problem question. If you want that same order in the conversation that follows, the SPIN Sales Conversation skill NL is the logical next step.

The closing question leans on Chris Voss, Never Split the Difference, 2016. Voss describes how a question someone can answer with no creates less pressure than a question that asks for yes. That is why no is always allowed to be a good answer in these emails, and why the follow-up email in the example ends with "should I let this go". That is not a trick but the removal of the pressure most acquisition emails get stuck in.

For reciprocity and social proof, Robert Cialdini's Influence, 1984, applies, and for attention his Pre-Suasion, 2016. The skill uses only the honest side of those: evidence you actually have, and attention you earn through relevance rather than a misleading subject line. If you want to work out the six principles separately for a text, that is what the Cialdini Persuasion skill is for.

The rule that you write in plain words is backed by Daniel Oppenheimer in Consequences of Erudite Vernacular Utilized Irrespective of Necessity, Applied Cognitive Psychology, 2006. Needlessly complicated language does not make the writer look smarter to the reader. For the questioning that comes before the email, The Mom Test by Rob Fitzpatrick, 2013, is on the list: talk about their work, not your idea. That same attitude sits in the Mom Test Coach skill.

Then the legal foundation, in broad outline, because the skill is not a lawyer and says so itself. The Dutch ban on spam sits in article 11.7 of the Dutch Telecommunications Act (Telecommunicatiewet) and applies not only to consumers but also to business recipients: the third paragraph makes an exception for contact details the recipient has themselves designated and published for this kind of message. The Dutch Authority for Consumers and Markets (ACM) supervises this. Source checked on 25 August 2026.

Every commercial email must include a working opt-out option, and a line at the bottom where the recipient can reply that they do not want to hear anything more counts as such an option, provided you also stick to it. The sender must be recognisable: a real name, a real company, a real address, and a subject line that covers the content.

For a business email address with a person's name in it, the GDPR also applies, so you must be able to explain where you got the address from and why you are approaching that person, and you must honour an objection. Exactly where the line sits with a general address of a legal entity is a question for your lawyer, not for a text file.

What the skill does with that broad outline is limited and concrete: it puts the recognisable sender block and the opt-out line into your email as standard, it refuses a misleading subject line, and it notes in the notes section what you said about where the address came from. What it never does is judge that your approach is allowed. That distinction is explicit in the file, and it is exactly the kind of honest boundary that is completely missing from most outreach skills on GitHub.

~/skills/ai-acquisition-email-writer/04-refusals[ok]
Section 04 · The limits

What the skill refuses

reg. A.004

The SKILL.md contains a list of twelve things the skill never does, and in cold acquisition that list matters at least as much as what it does do. An email with a trick in it does not just go wrong with this recipient: it gets forwarded to a colleague, it gets read aloud in a meeting, and your name stays attached to it. Here is how those rules play out: each row is a request you might make, with the response the skill gives according to its own instructions.

FIG.04: refusals.log[####-] sheet 4/5
refusals.log12 fixed rules from the SKILL.md
I have 180 addresses from a list, write one email I can send to all of themREFUSEDNo email without a concrete observation about the recipient. A list of 180 companies says nothing about any one of those companies. What you get back are three questions per company, and with that the skill will write ten emails that do not resemble each other.
put Re: in the subject line, it gets opened more often that wayREFUSEDNo misleading subject lines. Re:, Fwd:, your enquiry, our agreement, invoice: it suggests a conversation that never happened. A recognisable sender with a subject line that covers the content is not a style choice under Dutch law.
mention that Peter van Dijk referred me, that opens doorsREFUSEDNo invented mutual acquaintance, referral or previous contact. If the recipient calls that person, you lose more than a client. If you genuinely know them and are allowed to use their name, say so and the skill will put it in.
say I only have two spots left this month, that creates pressureREFUSEDNo false urgency or invented scarcity. Any deadline that cannot be shown to be real gets removed by the urgency test. If you genuinely only have two spots, that is allowed to stay in, provided it is true.
add that we get 30 per cent more conversion on averageREFUSEDNo invented figures, client names, results or accreditations. Whatever you have not supplied does not go in the email. If a number you are allowed to mention is missing, [FILL IN] appears and the skill asks you for it.
write that I went through their whole websiteREFUSEDNo claim that you looked at, read or tried something if you have not said so. This is the most common lie in cold emails, and it shows the moment the recipient asks a follow-up question.
where can I buy good email addresses of companies in my regionREFERRED ONNo advice on buying or scraping address lists. The skill does not help compile these and points you to your own lawyer for what is allowed. Its own answer is simpler: pick ten companies you actually want to work for.
put two questions in, that raises the chance of a replyCORRECTEDNever more than one question and never two next steps. Two questions means having to think about which one to answer first, and that is exactly the moment an email gets left unanswered. The question test catches this.
just leave out that opt-out line, it reads so formalCORRECTEDNo email without a recognisable sender block and opt-out line. The skill writes it in plain language so it does not sound formal, but it does not leave it out. This block never gets dropped.
he said no, write an email where I try again anywayREFUSEDDo not continue after a no, and no third email in the sequence. After a no, the skill writes two sentences to say thanks and nothing more. After two unanswered emails the sequence stops, and it notes that too.
am I actually even allowed to email this address under the lawREFERRED ONNo verdict on whether your approach is legally allowed. The skill gives the broad outline around the opt-out option, recognisable sender and the GDPR, and points you to your own lawyer and the ACM's explanation for the details.
just give me the email, I do not need that check blockCORRECTEDThe check block never gets dropped. It sits under your email and does not go to the recipient, so it costs you nothing. Without that block you would not know what the skill changed in your draft, and that is exactly why it exists.
$ unzip ai-acquisitiemail-schrijver-skill-voor-claude.zip -d ~/.claude/skills/
DOC.05

Installing in Claude Code, Claude.ai or Codex

The zip contains one folder, ai-acquisition-email-writer, with the SKILL.md inside it. Installing is simply 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. So you are not downloading a Claude file but a working instruction any modern AI assistant can read.

CLAUDE CODE
  1. Unzip it into ~/.claude/skills/ (or .claude/skills/ in your project).
  2. Claude then recognises the skill by itself as soon as you start talking about acquisition, cold emails or prospects.
  3. You can also call it directly, with /ai-acquisitiemail-schrijver.
CLAUDE.AI
  1. Go to Customize and then Skills.
  2. Upload the zip there as a skill.
  3. Or paste the contents of SKILL.md into the project instructions of a Project.
CODEX
  1. Open AGENTS.md in your repo.
  2. Paste the contents of SKILL.md into it, or place SKILL.md alongside it as a separate file and refer to it from AGENTS.md.
  3. Codex reads that along at every session.

After that, using it is simple: name the company, say what you know about that company and where you saw it, and say what you do. If something is missing, it asks you itself, at most three things at a time. If you already have an email you have written and just want to know what is wrong with it, paste it in full: the second mode assesses it on five points and hands back a rewritten version. If you get stuck installing it, the full step-by-step plan per environment is at installing Claude skills NL, and the broader explanation of working with AI is in the knowledge base.

$ cat 06-wanneer-wel-en-niet.md
DOC.06

When to use it, and when not to

It works best for a handful of companies you genuinely would want to work for. The file also names six variants where the approach shifts. For the first cold email, the full structure applies, with no price. After a trade show, a drinks event or a webinar, the observation is the conversation you had, and the opening may be more direct, because that email is no longer cold. If there is an inbound signal such as a download or an enquiry, you start with what the recipient did rather than what you offer, in three sentences and a question.

At a larger organisation you write to a person and not a department, and then the closing question is allowed, as the only exception, to expect a name as the answer: does this sit with you or with someone else. A no is followed by a two-sentence thank you and nothing further. And a restart email after months of silence is only allowed with a new observation, because without a new trigger it is simply a third email.

There are also situations where you are better off leaving it alone, and the file is brief about that. Mass emails to a bought or scraped address list are out. Newsletters and mailings to existing clients are also out of scope, because there the relationship is different, and so are the rules. A job application letter is not for this either, however tempting the resemblance. The proposal that follows a good conversation belongs with the Offer Writer, not here. And it is not fit for use as a legal check on your email policy.

One more honest limit: the skill does not fetch addresses, it does not look anything up and it does not browse websites. Everything it writes about the recipient comes from what you give it. That feels like a limitation and it is exactly why its emails are credible: there is not a single sentence in them you cannot defend if the recipient pushes back. The research stays human work, and for ten companies that is about an hour.

If your question sits closer to the conversation than to the email, the SPIN Sales Conversation skill NL picks up from there. If you first need to get clear on why a recipient would respond at all, the Jobs to be Done Analyst is the better starting point. If the conversation that follows needs to be recorded, the Intake Summary skill does that, and once it comes down to price, the Negotiator takes over.

~/skills/ai-acquisition-email-writer/05-upgrade-path[ok]
Section 05 · From skill to employee

Run it yourself, or have it run for you

reg. A.005

This skill is the free do-it-yourself version of work we also deliver as a service. What you get stays complete and without any catches, but be aware of what a skill is: it teaches your AI how to do something, while every new session starts empty. It is not the engine and it is not the memory. You find the observation, you prompt, you keep track of who you emailed when and who said no, you check. Anyone who wants that differently has two next steps: hand off the engine, or sort out the memory.

$ cat from-skill-to-employee.mdthree steps, same work
upgrade-path.shfree · employee · brain
STEP 1 · FREE
where you are now
The skill: you are the engine You run the Acquisition Email Writer yourself in Claude, Codex or Cursor. Costs nothing, works today, and you keep it entirely in your own hands: no trial period, no locked off parts. The limit is your own time: an email only appears when you ask for one, you find the observation yourself, and you keep your own list of who you emailed when. $ claude --skill ai-acquisition-email-writer · €0 · you prompt, you check
STEP 2 · SERVICE
have it prepared
The Sales employee: it is ready without you prompting You have the conversation, that is not something you outsource. The work around it, though: preparing the companies whose turn it is this week, looking up the trigger per company, the draft emails sitting ready, the follow-up that comes back on day five, and the list of who said no and therefore gets nothing more. That is what an AI employee sets up every day, with human final approval as a fixed part of it: output stays a draft until a person gives the go-ahead. We deliver that through Mansotti, the company TheSEO trades as, which today builds three employees: the Quote employee, the Sales employee and the Reporting employee. For this work the Sales employee is the logical one, because prospects and follow-up are exactly its territory. from €950 per month · a person gives the go-ahead, always
STEP 3 · BRAIN
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. In acquisition that is the difference between a one-off email and a memory: who you approached in March, what they replied, who said no and is therefore left alone, and which observation you used at the time. What that delivers in practice, from the plans to your first week, you can read at Jarvis itself. entry plan Brain Start: €9 per month incl. VAT · pay for your brain, not per AI question
# no marketing trick: step 1 stays free and complete. The next steps are there for anyone who wants to hand this work off.
~/skills/ai-acquisition-email-writer/06-jarvis[ok]
Section 06 · The brain

What Jarvis actually delivers

reg. A.006

Step 3 deserves more than a paragraph, because this is the difference between a clever chat and a system you can build on. Jarvis is the organisation brain: it remembers what your AIs need to know, divides up the work and keeps track of what happened. In acquisition that stands out even more. A chat window does not know you already emailed this company last quarter, does not know the director said no at the time, and does not know which observation you used then. That is exactly the knowledge you need so you do not land the same email with the same recipient twice. You notice it first at the start of a new session.

FIG.05: What a session gets back from the brain[#####] sheet 5/5
jarvis · organisation brain● sync
$jarvis recall "acquisition north brabant" # schematic example
[core]every cold email starts with an observation and where it was seen, never with the offer
[core]after two unanswered emails the sequence stops · after a no the address goes on the do-not-approach list
[task]four draft emails set up for this week · waiting for human approval
[decision]no prices in a first email: recorded after review, applies across all channels
[log]previous session: approached this company in March, director replied later, a person adjusted the opening
[ok]context loaded · this session does not start empty
every task moves through the brain like this: recordedcontext setdelegatedhuman approvalstored · the full trail is at jarvis/werking NL
context.retained Your next session does not start over Today you say which companies you are approaching and with which trigger, and tomorrow a standalone chat knows nothing about that any more. With Jarvis every session starts with the same projects, core knowledge and earlier decisions, as in FIG.05: recall first, only then write.
ai.connected ChatGPT, Claude and Codex, one source Every connected AI works from the same core knowledge and agreements. What you record and approve in one tool is used by the others too. You do not explain your approach three times, and no three different versions of your story spring up side by side.
tasks.tracked Tasks scheduled, tracked, reported done A task is recorded with a goal and a deadline, picked up by the right agent and reported done with the result attached. Exactly what a follow-up email on day five needs so it does not sit there until you happen to think of it again.
everything.logged Everything logged and open to view Every step leaves a checkable trail: who asked what, which sources were used, which agent worked on it and who gave approval. In acquisition that is also your answer to the question of where you got an address from.
human.approval Nothing goes out without approval AI prepares, a person decides. Output stays a draft until someone approves it, and only approved knowledge goes back into the brain. For emails that go out to the world, that is not a luxury but the entire condition.
brain.isolated Client brain isolated If you work for multiple clients, the knowledge stays strictly separated per brain. What you learn about prospects and agreements for one does not leak into the work for another. In sales that is the difference between trust and a problem.
# THE HONEST PROOF · NOT A DEMO

We have been running on this system ourselves for months, including for our own acquisition. Every agent session, every task and every decision gets logged in it and can be read back. A new session therefore does not start blank: it first pulls up the recorded decisions, the running projects and the latest changes, and carries on where the previous one stopped. So we are not describing a promise but the way of working we ourselves work in every day.

$ cat pricing.mdpay for your brain, not per AI question
Brain Start €9 /mo incl. VAT 1 organisation brain · 1 user · 1 AI employee
Brain Solo €29 /mo incl. VAT 1 organisation brain · 1 user · 3 AI employees
Brain Team €99 /mo incl. VAT 1 organisation brain · 5 users · 10 AI employees
Brain Business €249 /mo incl. VAT 3 organisation brains · 20 users · 50 AI employees

See all four plans at jarvis/prijzen NL. Through the waiting list NL you only pass on your preferred plan, without obligation. That does not yet create an account, an order or a payment obligation. We discuss business custom work first.

~/skills/ai-acquisition-email-writer/07-related-skills[ok]
Section 07 · The conversation map

The skills around it

reg. A.007

An acquisition email is one step in a sequence. These skills from the same library cover the pieces before and after it: working out why someone would reply, the conversation that follows, and the other channels in which you tell the same story.

$ claude --interactief # seven questions, seven answers
DOC.07 · FAQ

Frequently asked questions

What does the Acquisition Email Writer skill cost?

Nothing. The skill is free, the file itself states the MIT licence, and you do not have to create an account or leave an email address. You download a 9.3 KB zip containing a folder and a single file, SKILL.md, and that is the complete skill. There is no paid version and no sales email follows, which is the least you can expect from a skill about acquisition emails.

Does this skill also work in Codex, Cursor or Gemini CLI?

Yes. SKILL.md has been an open standard since December 2025, so the same file also 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. Putting the contents of SKILL.md into your AGENTS.md still works too. The instructions are plain readable text, so any AI assistant that accepts instruction files can handle it.

Why does the skill refuse to write if I do not know anything about the recipient?

Because then there is nothing to write that concerns that recipient. The skill tests your observation against three requirements: specific, traceable and useful. If it fails there, the opening would fit just as well on any other company in the sector, and that is exactly the email that does not get answered. Instead of an email you get three questions back that let you find the observation yourself. That costs you a few minutes per company, and that is also why this skill is not suitable for a list of two hundred addresses.

Am I allowed to just send a cold business email in the Netherlands?

The file only gives the broad outline on that and points to your own lawyer for the details. That broad outline: the Dutch ban on spam sits in the Telecommunications Act and also applies to business recipients, the ACM supervises this, every commercial email needs a working opt-out option, and the sender must be recognisable with a subject line that covers the content. For an address with a person's name in it, the GDPR also applies. The skill puts those elements into your email as standard, but it never judges that your approach is allowed.

What exactly is that self-check?

A fixed step in which the skill attacks its own draft with eleven tests before delivering anything. The best known is the copy test: replace the recipient's company name with any other and see if the email still holds up. If it does, the opening is a template and the skill delivers questions instead of an email. It also counts I-forms against you-forms, checks whether the closing question can be answered in one word, and removes urgency that is not real. What has been corrected is listed in the check block under your email.

Does the skill also write a follow-up sequence?

It delivers one follow-up email with a different angle, not the same text with a reminder stuck on top. And that is where it stops. After two unanswered emails the sequence stops, and the skill notes that too. After a no it writes nothing further at all, except two sentences to say thanks. Anyone who wants a sequence of seven emails is at the wrong address with this file.

When should I specifically not use this skill?

Not for mass emails to a bought or scraped address list, because the skill will not start on that. Not for newsletters and mailings to existing clients, because that is a different conversation. Not for job application letters. Not for the proposal after a good conversation, the Offer Writer skill is for that. And not as a legal check on your email policy, because that is work for a lawyer and not for a text file.

$ cat 08-en-nu.md
DOC.08

The trigger first, then the email

Anyone who picks ten companies and spends three minutes per company on the question of what stands out to them there writes better emails than anyone who emails two hundred. That is the whole argument behind this skill, and you can apply it even without downloading anything.

The same goes, by the way, for the other side of the coin: while you are sending emails, people are actively searching for what you do. The free SEO scan shows in a few seconds how your site is doing technically, and we explain there what an outsourced marketing department NL can do for you. If you want your team to learn to work with skills like this, we do that in the AI training NL. And if you want to talk further about what else AI can do for your business, from individual skills to full automation, we just do that in a conversation.

Section 08 · Next stepreachable 24/7
Book a call NL
$ whoami
Gianluca, founder
Written by GianlucaFounder. Has been building visibility for Dutch companies since 2017, in Google and in AI answers. More about the institute.