Flowscale
FOR AGENCY OWNERS WHO SEND (OR SHOULD BE SENDING) · ENGINE #2 OF THE FIELD KIT · RUNS IN YOUR OWN CLAUDE

// THE OUTREACH ENGINE · INSTALL IT, OWN IT

Research in.
Ready-to-send approach out.

30 cold emails a week is the floor for an agency that wants to grow. Most owners send 4, because every one starts with a blank page. This engine kills the blank page: 3 fact-cited openers and a 3-email sequence in YOUR voice, for any prospect, in minutes. $47 USD once.

Read a real run ↓

Runs entirely in YOUR Claude account · writes, never sends · keep-the-files guarantee

// THE BLANK PAGE TAX

Your prospects can smell a template in two seconds.

You know the math: consistent outbound is the difference between a pipeline and a prayer. And you still don't send, because writing one good cold email takes 25 minutes, and writing a bad one costs a prospect forever.

So you tried the tools. "Personalization at scale." {first_name}, {company}, a scraped compliment about a LinkedIn post. Reply rates fell anyway, because the receiving end has learned exactly what manufactured familiarity looks like.

The emails that still get replies have one thing in common, and it isn't the template. It's that the sender visibly did the homework.

// THE WRONG FIX

You don't need better templates. You need outreach built on evidence.

A real observed gap on their site. A review quoted verbatim. A competitor's ad count from this morning. When the first line of an email contains a fact the prospect recognizes as true, everything after it gets read differently.

That kind of outreach was always too slow to do at volume, because every email needed the research first and then the writing. The Outreach Engine does the writing half in minutes, from research it can cite. Pair it with the Dossier Engine and both halves are machines.

// WHAT IT ACTUALLY IS

A working skill with a built-in skeptic. Not a swipe file.

You type: "Run cold outreach on prospects/cedar-park-family-orthodontics.md". The engine reads the research, then works like a copy chief:

1 · EXTRACT THE AMMUNITION

The angle, the top pains with their evidence, the trigger, the objection, the decision-maker. From a dossier file or your pasted notes.

2 · PICK FROM 20 OPENER FORMULAS

Each keyed to evidence with a "use when" condition. The Review Echo doesn't fire unless there's a review to echo.

3 · WRITE THE SEQUENCE

Day 0 opener, Day 3 new-value follow-up (never "just bumping"), Day 7 clean breakup. Subject lines included. Under 120 words each.

4 · AUDIT THE PERSONALIZATION

Every personal fragment must trace to a real observed fact. Fake familiarity gets struck, "loved your recent post" doesn't survive.

5 · APPLY YOUR VOICE

You give it two lines you actually wrote, plus your greeting, sign-off, and banned phrases. It calibrates. Not AI voice, your voice.

6 · FACE THE SKEPTIC

A built-in reviewer grades every touch like a busy owner who deletes 29 of these a week: SEND, FIX, or REWRITE, with the reason.

TWO RULES IT CANNOT BREAK

It only uses proof points YOU gave it; it will not invent a result to make an email stronger. And it never sends anything, it writes to a file, you send from your own tools. Your list, your domain, your call.

// THE OUTPUT, NOT A PROMISE

A real run, typeset for the page.

Run on the sample dossier (Cedar Park Family Orthodontics, the fictional representative company from the Dossier Engine) with the kit's worked-example ICP and voice. Your runs use YOUR offer, YOUR proof points, YOUR voice.

outreach/cedar-park-family-orthodontics.md · angle: boosted-post leak + second location
EMAIL 1 (DAY 0) subjects: "your boosted posts" / "the second location math" / "Cedar Park's ad spend" Hi Dr. Calloway, Your Meta ads are running as boosted posts. For a two-location ortho practice that usually means paying 2-3x more per new-patient lead than the same spend optimized. From the outside I can also see the gap after the click: form fills with no fast follow-up, and a second location that needs chairs filled on a schedule referrals don't keep. I cut new-patient cost-per-lead to $21 against the $35-75 band for a practice with this exact setup. Want the 5-minute teardown of where yours is leaking? It's yours either way. Best, Sam Northbeam Digital EMAIL 2 (DAY 3) subject: "wrote it up anyway" Hi Dr. Calloway, I went ahead and wrote the teardown: the boosted-post leak, the form-to-phone gap, and the 30-day plan for filling the new location's chairs. Two pages, no call required, no strings on it. Want me to send it over? Best, Sam EMAIL 3 (DAY 7) subject: "closing the file" Hi Dr. Calloway, Closing the file on this unless it's useful. If the new location's patient flow becomes a now-problem, the teardown is yours, one email away. Good luck with the opening, genuinely. Best, Sam

THE SKEPTIC'S VERDICT (SHIPS WITH EVERY RUN)

TOUCH 1: SEND. Opens on their problem, every claim traceable, one small ask. Reply test: "It names my actual setup and asks for interest, not my calendar. I'd answer." TOUCH 2: SEND. Adds the artifact instead of bumping. TOUCH 3: SEND. Breakup with the door open. Note: highest reply probability in the set; do not skip it.

Note the sequence never says "just following up." The reviewer kills that phrase on sight, along with "I hope this email finds you well."

// WHAT SHIPS

Every file, annotated. No mystery box.

my-agency/ · THE OUTREACH ENGINE
├── icp.md <- your offer + real proof points (shared with the Dossier Engine)
├── voice-guide.md <- how YOU write; two real lines calibrate the engine
├── prospects/ <- research in (dossiers or your own notes)
├── outreach/ <- sequences out, one file per company
└── .claude/
    ├── agents/reply-reviewer.md <- the built-in skeptic
    └── skills/cold-outreach/
        ├── SKILL.md <- the engine: ammunition → angle → sequence → review
        ├── opener-formulas.md <- 20 patterns, each keyed to evidence
        └── sequence-framework.md <- 3-touch + 5-touch structures, reply handling

Plus the install walkthrough on the delivery page, written for someone who has never opened a terminal. Full source in copy-paste blocks AND a zip. Edit anything; it's yours.

// WHAT YOU ACTUALLY NEED · READ THIS BEFORE BUYING

The honest requirements.

Why it works this way: I never resell Claude access, and your prospect list and copy never touch my servers. The engine runs entirely inside YOUR account, and it never sends a thing, you stay the sender, on your own domain, under your own rules.

// HOW IT WORKS

Three steps. Then outbound stops being the thing you skip.

01

Install

Unzip into your agency folder, install Claude Code, log in. About 5 minutes with the walkthrough.

02

Teach it your voice

Fill voice-guide.md with two real lines you wrote, your sign-off, your banned phrases. 10 minutes, once.

03

Point it at research

A dossier file or pasted notes. Minutes later: openers, sequence, verdicts, in outreach/. You send.

THE SYSTEM BEHIND IT

This is the outreach discipline from the same system my own agency runs on, the one that's pulled leads at $21 against a $35-75 USD band and inbound calls at $5-6. Pain-first isn't a slogan in it, it's an enforced rule: if the first line isn't about THEIR problem, the reviewer fails the draft.

// THE MATH

$47 USD. Once.

A freelance copywriter charges more than $47 for ONE sequence, once. Subscription outreach tools bill you monthly forever and still write from templates. This is $47, one time, for unlimited prospects, with your voice and your facts baked in.

The guarantee: if you can't get it running, I will personally get you running, or refund the $47. You keep the files either way.

// STRAIGHT ANSWERS

Questions a smart buyer asks.

Do I need the Dossier Engine first?+

No. Feed this engine your own notes and it works: it does a quick web pass of its own before writing, and labels what it found. But the pairing is the point: dossier in, approach out, sharing one icp.md and one prospects/ folder. Research-built openers are simply better ammunition than notes-built ones.

Will it actually sound like me?+

You give it two real lines from emails you actually sent, plus your greeting, sign-off, sentence-length preference, and banned phrases. It calibrates on those, and the reviewer checks every draft against them. The first run might need one nudge ("shorter," "drier"); edit the voice guide once and it sticks. If you leave the guide empty it writes in clean conservative defaults and tells you so.

Is this a spam machine?+

The opposite, structurally. It writes researched, one-at-a-time outreach and never sends anything, you stay the sender. It bakes in the hygiene rules (real name and business, honor every "no" permanently, your local email law governs sending) and refuses fake familiarity. If you want a blast cannon for purchased lists, this is the wrong product and I'd rather you not buy it.

I've never opened a terminal in my life.+

The walkthrough was written for you: install one app, paste one command, log in with the Claude account you already have. About 25 minutes once, most of it filling in your voice. Genuinely stuck? Email me, I get you running or refund you. That's the guarantee.

What does running it cost me?+

The Claude Pro plan, $20/mo USD, paid to Anthropic, the same plan that runs the Dossier Engine and everything else Claude does for your business. No per-sequence fee, no subscription to me. Pro's limits comfortably cover normal daily outbound writing; heavy power users can upgrade to Anthropic's Max plan, but start on Pro.

Does it write LinkedIn DMs too?+

Yes. Every run includes an alternate-channel opener (usually a DM: under 60 words, no link in the first message, colleague tone, not campaign tone), and the framework covers how DM and email touches interact, if the DM gets a reply, the email sequence stops.

What if Anthropic changes pricing or terms?+

Then your costs change with them, like every platform your agency already stands on. What you own outright is the procedure, the formulas, the framework, and every sequence the engine ever writes, plain text on your machine, nobody can re-price it.

// FOR THE SCROLLERS

The short version: the Outreach Engine is a working skill you install in your own Claude account. Feed it a researched prospect (a Dossier Engine dossier, or your own notes) and it writes 3 openers chosen from 20 evidence-keyed formulas plus a 3-email sequence in YOUR voice, then a built-in skeptic grades every touch like a recipient who deletes 29 a week. Pain-first enforced, no invented numbers, never auto-sends. You need Claude Pro ($20/mo USD, from Anthropic) and about 25 minutes of setup. $47 USD once, unlimited prospects, keep-the-files guarantee.

// ONE ROOM OF A BIGGER HOUSE

The approach is only as sharp as the research feeding it.

The Dossier Engine runs the 8-pass recon that fills your prospects/ folder with dossiers: fit score, live pain, triggers, objections. This engine turns those into the approach. Two machines, one brain, and the bigger system they come from ships one engine at a time in the Field Kit.