Skills · Brand Story · Open Source

The storytelling skill my own marketing runs on.

Brand Story is the Claude Code skill behind the content layer of this one-person marketing function. It plans and writes LinkedIn and X content the way an operator would: research first, proof required, and a hard rule that kills any post that can't back its claim.

It's free, open source, and brand-agnostic — the same engine, pointed at your brand instead of mine.

Get the skill on GitHub →

Part of Proofstack · MIT license

The Receipts Protocol

"Most content tools optimize how things are said. This one gates what's allowed to be claimed."

AI can produce unlimited confident marketing copy — which is exactly the problem. The scarce resource isn't words anymore; it's claims a reader can check. The protocol enforces that scarcity: every post traces back to a live conversation it's entering and a receipt it's carrying. If either is missing, it doesn't ship.

How it works, in five moves

01

Listen before writing

Every planning run opens with live social listening — X, Reddit, and Instagram via MCP tools, LinkedIn via scraping actors. The output is a Signal Brief: what's hot, what's saturated, what's unclaimed, and the audience's own words. Every signal carries a source URL, so the research is checkable the same way the content is.

02

The five-line brief

No draft exists before the brief. Five lines force the discipline most content skips: a named reader, a live misconception, one claim, one receipt, one action. The receipt line is the kill switch.

03

The receipt gate

The rule that gives the protocol its name: a claim without a verifiable number, named experience, or public artifact behind it does not ship — regardless of how well it would score. LLMs invent statistics fluently. This is the structural defense.

04

Platform-native drafting

LinkedIn and X get different drafts, always — different hooks, different structure, different mechanics. Five story arcs (Receipt, Reckoning, Build Log, Confession, Teardown) rotate across four content pillars so no two weeks read the same.

05

The feedback loop

Every week's results are logged — pillar, arc, hook pattern, engagement quality. The next planning run reads the log before it plans. The system gets sharper because it remembers what missed.

The five-line brief

Before the skill writes a word, it writes five lines. It's the same brief I gave human teams for a decade — compressed to the version an AI system can't wriggle out of.

01 · WHO

One named reader — not a segment. "Series A founder about to hire a CMO," not "B2B decision-makers."

02 · WHAT

The thing they believe today that's wrong. Every good post corrects a live misconception.

03 · CLAIM

The one thing this post proves. One. Posts that prove two things prove nothing.

04 · RECEIPT

The number, story, or public artifact that proves it. If this line is empty, the post dies. No exceptions.

05 · ACTION

What the reader does in the 60 seconds after reading — something to do, feel, or steal.

Why trust it

This isn't a template. It's an organ of a working system.

Most marketing prompts on the internet have never marketed anything. This skill is extracted from the system that runs my actual marketing function — the same discipline documented with live numbers on The System: the hours, the costs, the output, updated as they change.

The public version is brand-agnostic: it reads your voice, audience, and positioning from your own brand context file, and the repo ships with a worked example showing a full research-to-draft run. If it produces something my own review gate would kill, that's a bug — file it.

Newsletter

The Autonomous Marketer

The decade of marketing I wish someone had written down. Sent when I have something real to say — not on a schedule.

Subscribe on Substack

Free. Unsubscribe anytime.