The Shelf / Strategy / Positioning Market Map
Positioning Market Map
One page that says who buys this, who else they'd pick, and why you win. No code.
The job: you proved people want it. Now you need one page a stranger could read and get it — who buys this, what they do today instead, how the money arrives, and why you win. Hand the file to your agent, get back one document, docs/positioning.md. No code.
Most positioning chats with an AI go the same way. You describe your product. It tells you the idea is great. You walk away with "for busy professionals who want to save time." That page is worth nothing. This is the gear that turns your agent into the annoying strategy partner instead — the one who asks who exactly buys, and won't take "small business owners" for an answer.
What it makes your agent do
- Check you have proof first. No evidence anyone wants this, no session. It won't polish a guess.
- Ask one question at a time until your buyer is a real person — their day, who else signs off, a line they'd actually say. Not a category.
- Line up four kinds of competitor, including the one everybody skips: doing nothing. Most products don't lose to a rival tool. They lose to "we're fine as we are."
- Make you pick how you charge — the price to anchor on, rough costs, and the exact route to your first dollar.
- Shrink the market to one slice: this industry, this job, this stage. Small enough to win inside a year.
- Write the one sentence that says why you win. The way you do it has to be specific, the "unlike them" has to come from the competitor list you just built, and the claim has to be one you could be proven wrong about. A buzzword is not a mechanism, and the skill says so.
Who it's for
Founders and solo builders stuck between "people want this" and "start building." Anyone who runs agents and wants every project held to the same standard. Consultants who do this work for clients.
Why not a free directory download
Free skill folders are full of unchecked files that say "act as a marketing expert." This one has a real process in it — hard stops, answers it rejects, choices it forces, a rule against flattering you — sharpened across a working multi-product portfolio. And it's maintained: as the process gets sharper, you get the update. Read the field report below and see the exact deliverable before you pay.
Take it off the shelf, drop it in your skills folder, run it on your project tonight.
FIELD REPORT real output, not a promise
What the skill actually forces, excerpted from its own working material: the niche drill it runs, and the deliverable skeleton it fills in. No sample company invented — this is the gear itself, unpacked.
The niche drill (verbatim from the skill)
A niche is the intersection of vertical x use case x stage. The skill grades answers like this:
- ❌ "Marketing teams" — not a niche
- ❌ "B2B SaaS marketing teams" — still too broad
- ✅ "Series-B B2B SaaS marketing teams running outbound on Apollo with <5 SDRs" — a niche
If your answer looks like the first two, the agent keeps asking. That's the point.
The competitor matrix it requires — all four columns
| Column | What it captures | Why it matters |
|---|---|---|
| Status quo | The workaround used today (spreadsheet + email, manual process, an agency) | The real competitor. Most products lose to "do nothing differently." |
| Direct | Tools solving the same problem the same way | Pricing anchors, parity expectations |
| Adjacent | Neighboring tools that could expand into yours | Future threat |
| Outsider | A weird angle — another industry's solution, open source, a services player | Where disruption usually comes from |
Each entry needs: name, one-liner, pricing anchor, what they get right, what they get wrong, where they're vulnerable. Then one sentence naming the vacuum all four columns leave open.
The UVP form it will not let you dodge
For [the named ICP], who [pain, from evidence], we are [category], that [unique mechanism], unlike [an entry from the matrix — no invented strawman], we [the wedge difference — falsifiable, measurable, not a vibe].
Rules enforced: "AI-powered" is rejected as a mechanism; the "unlike" must exist in the matrix; the agent must state whether the final UVP is sharp or fuzzy and why — "nice positioning" is banned output.
The deliverable
One file, docs/positioning.md: upstream read, ICP one-pager, the four-column matrix with the vacuum named, business model with pricing anchor + path to dollar #1, the niche sentence with a defensibility note, the UVP, and open questions. Draft-stamped, ready for a revise loop.
SERVICE RECORD living gear — updated as the factory learns
v1.0.0 — 2026-07-17
First issue. Ported from the factory's internal skill: sanitized for general use, methodology intact, field report captured from a real run.
Every update ships free to owners — your locker always serves the latest version.
QUESTIONS
Does this work in Claude Code, claude.ai, and Codex?›
Yes. It's a plain text file. Claude Code: put it in your ~/.claude/skills/ folder. claude.ai: upload it to a Project. Codex: put it in a skills folder, or name the file in your AGENTS.md. The steps are in the QUICKSTART.
Does it write any code or build anything?›
No. It produces one document and nothing else. This is the thinking you do before you build.
What do I need before running it?›
Proof somebody wants this — customer quotes, reviews, sales, or your validation notes. The skill stops cold if the demand is still a guess, unless you tell it to carry on anyway.
How is this different from just asking the AI about positioning?›
Ask it plainly and it agrees with you. This makes it go one question at a time, refuse category answers, pull the 'unlike them' line from the competitor list you actually built, and demand a claim you could be proven wrong about. The discipline is what you're buying.