The 7 Claude skills we'd build first

Hey there,

David here.

If we started a GTM team from zero tomorrow, there are seven Claude skills we would build first.

I shared the list on LinkedIn this week. Here I want to go a level deeper. What each one actually does, and why we build them in this order.

First, what we mean by a skill. A skill is a written set of instructions for one job. It takes something a person does by hand and has Claude do it the same way every time. That last part is what makes it useful. The output on the hundredth account looks like the output on the first.

Here they are, in the order we build them.

01Who hears from you, and what they read

1. Account scoring: This one defines the scoring model before it scores anything. Hard disqualifiers gate the list. Weighted fit ranks it. Time-decayed signals are tracked separately, and each tier band comes with its own action. Every dimension names the field it reads, so nothing gets scored on data the enrichment cannot fill. Then it scores every account in the list and gives a one-line reason for the tier.

2. Account research: Deep research on one named account at a time. It writes a company brief and a person brief, explains why now, and attaches a confidence score. The confidence score tells you how much to trust the brief before anyone writes to that person.

3. Copy: This writes the cold email sequence from three inputs: the offer, the proof and the voice. It runs from the subject line and first line through to the follow-ups, then scores the copy before sign-off. It runs before any lead gets personalised, so the personalisation has a finished sequence to fit into.

4. Personalisation: This writes the per-lead line from a real signal. We calibrate it on one lead first, then batches of ten. It is locked after two rounds come back with no edits. Only then does it run across the list. Every line is tagged signal or fallback. That shows how much of the list had something real behind it.

Skills 1 to 4 decide who hears from you and what they read.

02What comes back

5. Reply handling: Every reply gets a tag: super-interested, interested, question, not now, not interested, out of office, wrong person, unsubscribe or spam. The skill drafts the response, and a human approves it before anything goes out.

6. Weekly report: It pulls the LinkedIn numbers from HeyReach and the email numbers from EmailBison. Then it writes the client update, with week-over-week and month-over-month trends. It stages the report, and a person signs off before the client sees it.

7. Deep audit: This reads the whole engagement, start to finish. It shows which segments, angles and messages actually worked, and it ends on one ranked set of recommendations.

03Why the order holds

Each skill works on what the one before it produced. Copy needs a scored list to know who it is written for. Personalisation fills slots in a sequence that copy has already written and scored. The deep audit only has something to read once replies are tagged and the weeks are reported

Build them out of order and you end up personalising emails to accounts nobody scored.

Our best Claude guides for 2026

If you want to start on your own versions, we put our 10 best Claude guides on one page. Several are skill files you can install in a Claude Project today. The prospect research guide sits closest to skill 2. The cold email skill is a good place to start on skill 3.

04One small thing you can do this week

Pick the job on your team that gets redone by hand most often. Write down the steps the person takes, in order. Add what they read before they start and what a finished result looks like.

That page is the first draft of a skill. If two people on the team would write it differently, that is worth knowing before Claude runs it.

I am curious which of the seven your team still does by hand. Hit reply and tell me, I read every response.

PS. If you want these seven built into your GTM, grab a time with us here: