We built 52 Claude Code skills that run a full sales and marketing team. They've produced $7.83M in qualified pipeline for one client, with zero new SDR hires. It runs as six specialists, the same roles you'd staff a revenue team with: 1. Strategy: reads your site, works out who you can sell to, scores every account out of 100 2. Signals: tracks the right accounts and waits for the trigger. a raise, an SDR hire, a new CRO, a competitor swap 3. Data: pulls real phone numbers and emails for the accounts worth chasing 4. Copy: writes each message off the specific signal it caught, never a template 5. Execution: ships the emails, the LinkedIn touches, and the follow-ups 6. System: wires the other five together so one prompt runs the whole thing Each skill is a plain text file. Anyone on your team can open it and change how it works, no code. This is the exact system we run at @Frontal_AI right now. The full build and a live run are in the video. Comment SKILLS and we'll DM you the repo. Must be following.

pov: ai burning through sending domains like mixtapes

you can just ask claude to build a @clay workflow that: > pulls your linkedin engagers > enriches name, title and company > scores them 0 to 100 against your icp > finds the email for anyone above 70 > pushes them into instantly full workflow:

Me and Stefan started @heyreach_io from our bedrooms in Macedonia. Buggy af at first. Since then: $17M ARR and 7,000+ teams running their LinkedIn outreach on it. This is how you go from “Hey” to deals closed in 2026:

I drove $586K in closed revenue from LinkedIn Ads last year, on the same platform everyone swears is cringe. Here's what most teams do: They upload an ABM contact list, point the campaign at form submissions, and wait. The platform does exactly what it was asked. It goes and finds the people in that audience most likely to fill in a form. Technically they're leads, but they'd probably never become pipeline. Reason? Poor quality leads. 3 things decide lead quality (and can help you get better leads): > What your creative says > The audiences you build > What conversion event you optimise towards If you get the first two right, leave the third on form fills, the platform still learns the wrong lesson. So ask it for something else... When a deal closes in HubSpot, that closed deal goes back to LinkedIn as the conversion event. The platform goes and finds more people like the ones who bought. Most ads teams never hear what sales closed. I wire closed-won straight back into targeting. That handoff is where it breaks for most people. The ad account sits with one person, the CRM with another, and nobody owns the line between them.

if you're doing GTM, read this:

A top GTM engineer and a quant trader often show the same characteristics. The first sign is:

We sent the same 1,000 cold emails twice and tripled the deals on the second run. The first run followed the classic playbook and produced 20 replies, 5 calls, and 1 deal. The second run produced 3 deals, because that batch had already seen our content 3+ times before the email arrived, and someone who has seen you 3+ times converts 2 to 3x better on the exact same message. That test is why I stopped believing in "add another channel" as a growth strategy. After auditing pipelines at 310+ B2B companies, I can tell you the single biggest issue with most GTM motions is the lack of a proper order in choosing and running growth channels. Channel two gets added because channel one felt done, channel three because two was slow, and now there are three half-finished motions with no way to tell which one earned which deal. Here's the way to build your GTM motions so that they actually work: 1. Run one channel until reply rate sits flat for weeks with nothing left to change in the copy. A slow week doesn't count, check the trend across weeks. 2. Then add LinkedIn, posted from the founder's profile. Ours booked over 1,000 meetings in a single year while the company page did basically nothing. 3. When both flatten, wire the two together instead of reaching for a third. Track who's engaging with your content, treat anyone who's seen it 3+ times as warm, and route warm people into the email sequence first. 4. Once the connected system tops out, the next offer is already sitting inside your delivery. The deliverability setup we ran as part of the core service now sells on its own at $60K/mo. 5. Past $250K/mo, stop targeting titles and start targeting what just changed inside the account: funding, hiring spikes in the team you sell into, a new VP inside their first 90 days. That's where the 5 to 11% reply rates on our signal campaigns come from while everyone else sits at 3 to 5%. And if signal targeting sounds like overkill, one Fortune 500 subsidiary started hammering one of our resource pages one morning, over and over. We sent them a 73-second Loom the same day and it turned into a €200K+ contract. The whole playbook is in the article below (bookmark for later):

@itsalexvacca is building a full GTM engine LIVE today 🗓️ Starting with TAM, then turning it into campaigns inside the tools you already own, with one prompt. He’ll build the TAM in Claude Code, show the exact engine he runs for clients, and explain why he wouldn’t start a campaign from a generic list of 10,000 names. Live today at 12pm ET. Link to join in the reply below 👇

Not sure anyone wants to hear this, but letting an AI send your cold list on autopilot might be the worst pipeline call you can make right now. I watch people do it every few weeks. We've run outbound for 275+ companies, and about half the ones coming to us now show up with their domains already cooked. Same setup every time: they aimed an AI at a purchased list, hit go, and for a day or two it even looked like it was working. The replies dry up first. Your email lands next to ten other AI emails that read exactly the same, so nobody answers, and since the AI wrote all of it you can't even tell which part to fix. Just a dead pipeline and no thread to pull. The domains are the part that doesn't come back. They were never warmed for that volume, they end up on spam lists, and rebuilding one runs months. Some you just retire. @InstantlyAI looked at billions of emails and put the 2026 average reply rate at 3.43%. Our signal-triggered sends at @Frontal_AI hold 13%. If you really think about it, the AI writing the emails is the same on both sides, above. The gap is that ours only fire when a company actually does something, on infra that's kept healthy, with a human signing off before anything leaves. We call that GTM engineering. Devs already sat through this exact movie. Vibe coding felt like magic right up until code nobody read started breaking in prod, and the fix was spec-driven development: write the spec first, keep a human on it before it ships. Outbound's getting the same correction. The AI can write and send all day. Whether you're still landing in inboxes six months from now comes down to the system underneath it.

You can literally get your own employees posting on LinkedIn and turn it into a pipeline channel. We did: 27 new clients and $153K in new MRR in under 3 months. Getting 12 busy people to write consistently is the part that kills most of these programs. But ours got easier over time instead, because the incentives sit with the person posting and the friction is close to zero. Our content lead @mailankhong built and runs the whole thing at @Frontal_AI. 175K+ combined followers on LinkedIn, and the team wants to post more than when it started. Here's what's different: 1/ The followers and the DMs stay with the employee, not the company page. My instinct was that the company paying for the program should own the audience, and she runs it the exact opposite way. She's right. Nobody keeps posting for a logo past week 6. 2/ Voice capture happens in a 30-minute conversation. A Google Form asking "what is your tone?" doesn't get answered truthfully. The real voice only comes out when you talk. 3/ One editor reads every draft and sends it back in twenty minutes. A draft that sits in review for days doesn't get posted, and the person who wrote it usually stops drafting after that. The twenty-minute loop is what keeps 12 people writing at the same time. 4/ Three kinds of voices post in parallel, and each covers what the others can't. Founders carry the point of view. Operators carry the proof, because an AE or an engineer writing about the actual work is harder to dismiss than anything a brand account says. Customers bring the warm intros. Most companies run the founder tier, stop there, and wonder why everything reads like a press release. 5/ The program answers to pipeline. Paid spend only amplifies what already proved itself organically. And the first call changes when the buyer has spent two months reading the person selling to them: they arrive with most of the convincing already done, and the deal moves faster from there. By her own rule 1, the followers belong to the person doing the work. This whole system is her work. Follow @mailankhong if you're looking to grow your pipeline through content and create scalable employee-generated content engines.

that was an internal doc. was.

most cold email fails because of what it asks for: 30 minutes, from a stranger, in message one. a buyer needs 7 to 13 touches before they'll talk money. most agencies stop at three and start from zero every monday. we went from $30K to $550K+ a month without adding a single send. the flip: stop circling strangers with follow-ups, build the world they orbit instead. when they're ready, you're the obvious choice. And here's exactly how to build that system: > 1/ build a better list before a bigger one. job-title filters return more accounts than you could ever work, and about 5% of them are looking right now. intent signals cut it to a couple hundred where something just changed: raised funding in the last six months, hiring for marketing, founder posted about a conversion problem last week. 54 signals like this hold 5-11% reply rates vs ~2% cold. > 2/ give before you ask. swap the meeting ask for a video teardown of their website or six campaign ideas built for them. a click costs the reader nothing next to a 30-minute commitment, so far more people click. and the click is what the rest of the system runs on. > 3/ track the click. the moment someone views your resource they enter your retargeting for 30 days, your nurture list, and your CRM flags them high intent. one click = touches on four channels without another email sent. through email alone you're the annoying guy. through four channels you're just everywhere. > 4/ connect it. one CRM. email engagement in, site visits in, LinkedIn activity in, engaged list out as an ads audience. visited pricing, watched the video, opened three times = sales pinged now. same 1,000 emails, both ways: > old: around 20 replies, five calls, one deal. reset monday. > new: 30 to 50 qualified leads, over time 10+ calls and three deals, every one tracked and warming. close rate moves too, 20% to 30%, because by call time they've watched your teardown and seen you in their feed for weeks. ask for the meeting last. it's the most expensive thing in outbound.

We drove $887k in pipeline with 1:1 ABM ads. 61 accounts, 183 personalized ads, $33k spend. I'm giving away the Claude Code skill that runs it. Free. I've put more than $2M in ad spend through Claude Code, and this is the skill I reach for most. 1:1 ABM is the thing everyone agrees works and almost nobody runs. The work is why. 61 accounts means 61 audiences, 61 landing pages, 183 ads, and every single one of them different. So you either pay a team to build it for weeks or you pay five figures for software that does it. That's the part the skill kills. Here's what it does per account, without you ever opening Campaign Manager: > Audience. Builds a dedicated ad set for that one company, cut down by department and seniority so you're in front of the buying group and not the whole payroll. It also maps which of your accounts clear LinkedIn's 300-person floor, so you know what's even runnable before you build anything. > Creative. Generates the ads for that specific company, carrying their name. > Copy. Writes against that account instead of dropping a merge field into a template. > Landing page. Ships one page per account and wires every ad to the page built for it. The buyer clicks an ad with their company on it and lands somewhere about their business. Then it launches the whole set programmatically. 61 campaigns, 3 ads each, live. That is the entire job of a 1:1 ABM campaign, run from a terminal. I built it for client work and I'm crazy enough to have made it public. Comment "ABM" and I'll send it over. Follow me so I can dm you.

instead of buying lead lists, open claude code and do this (to find your best-fit leads for a fraction of the cost): 1/ drop an .env in the folder with five keys. @exaailabs and @linkup_platform for the research, ai ark to build the buyer list, lima data for verified emails, @instantlyai for sending. 2/ write a CLAUDE.md. your offer, your icp, who you sell to. it loads every session, so you stop re-explaining your business from scratch. 3/ write a list builder SKILL.md. your enrichment steps and your rules for what counts as a valid contact. 4/ point a copy agent at the scripts that already got replies. ours reads the ones out of 23m+ sends. give it nothing to read and you get slop. it can only copy what it's seen. that's the setup. then it's four prompts. 5/ research "please find companies in the us, vc backed, 50 to 200 employees, that have raised money recently. then research what changed for them after raising, and pull any signal we could use later in the copy to justify reaching out." the second half is the part almost nobody asks for. a company that raised ten months ago and is now opening sales roles is telling you something, and that's your first line. ask for it as an md file. that's fantastic context for your next prompt. 6/ list "using the list builder skill, find the head of sales or vp of sales at those six companies using ai ark, and validate the emails so we know they'll land." you're describing a seat, so it finds versions of it you'd never tick a box for. validation happens inside the same prompt, so you get a finished list after this stage. 7/ copy "read the copy agent and based on the skill, write three emails for those six prospects before we push to our instantly campaign." one line per prospect gets personalized and carries the signal from prompt 5. everything else is identical for everyone. more personalization than that doesn't get more replies. 8/ load "create a new campaign in the instantly account, name it recent funding series a series b head of sales, and push the six leads and the three emails for each prospect." you can get six prospects or even thousand using just the four prompts I added above. with volume, you have to scale your email infra. described the entire flow in the video below: (bookmark for later)

How to automate your ENTIRE outbound using Claude Code (full setup)

API-based GTM is here. You no longer need ten tabs open to set up an outbound campaign. With the right infra, Claude Code one-shots the entire process: > list building > lead scoring > email drafting > loading into your sequencer The whole thing, "done for you." Full walkthrough in the video below (bookmark for later):