Skip to content

Makes

Messaging house

Builds one core message, 3 or 4 pillars and the proof under each, in your buyers' own words, so every page, ad and email says the same thing.

Other tools

Add this skill to Claude

Claude takes skills as a ZIP file. Your download has started; if it didn't, use the button below.

  1. 1In Claude, open Customize, then Skills.
  2. 2Choose +, then Upload a skill, and pick messaging-house.zip.
  3. 3Ask for it by name, or let Claude use it when a task matches.

Skills need code execution turned on in Claude's settings. On a Team or Enterprise plan, an admin can share it with everyone.

Download the folder

Try it now, without installing

See the skill work on a made-up input before you add it. Nothing is installed, and you can swap in your own input.

Open in the Claude app

Opens a new chat in the Claude desktop app with the prompt filled in. Nothing is sent until you press enter.

Includes the whole SKILL.md and the sample input. Paste it into Claude, ChatGPT or any chat.

Influenced by
Message mining
Needs
Your positioning (alternatives, unique attributes, value, best fit and category)
Works with
Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md
Version
1.0.0 · Last checked · License MIT
Maintained by
Sabih AhmedCreative
SkillStrategy and messaging

Messaging house

Example output

Build the house from buyer doubts
DoubtPillarProof
Will lawyers use it?Works inside WordProduct docs
Will it catch our clauses?Your clauses, every timeProduct docs
Is it worth it?Faster first draftsProof needed

Builds one core message, 3 or 4 pillars and the proof under each, in your buyers' own words, so every page, ad and email says the same thing.

The skill

The whole skill, in plain text.

Nothing is hidden. Read it, change it to fit your team, or take the whole folder.

Scripts
Optional Python scripts for the counting. It works without them and says which way it ran.
Network
Opens web pages or your tools if your agent can. Works from pasted data if not.

3 test cases

messaging-house/SKILL.md

In the folder

  • messaging-house/SKILL.md
  • messaging-house/references/mining-guide.md
  • messaging-house/scripts/phrase_bank.py
---name: messaging-housedescription: >-  Builds a B2B SaaS company's messaging house: one core message, 3 or 4 pillars that each answer a different buyer  doubt, and the proof under each, worded from phrases mined line by line from interviews, calls and reviews, so  every page, ad and email says the same thing. Stands on the profile's positioning and buying moments, checks  itself with voice-check and claim-check, and writes the profile's Messaging house section. Returns READY, DRAFT  or BLOCKED. Use it after positioning, before a homepage, landing page or ad themes. Not for deciding positioning  (use positioning), for testing copy on buyers (use message-test) or for writing the page itself.license: MITcompatibility: >-  Works in Claude, ChatGPT, Codex, Cursor and any agent that reads SKILL.md. Uses a connected call recorder, CRM  notes or review exports if the agent has them, otherwise pasted text. Python 3 is optional, for mining phrases  and checking the house.metadata:  author: SHAKEUP  version: "1.0.0"  stage: make  plugin: strategy-messaging  homepage: https://www.shakeuphq.com/skills/messaging-house--- # Messaging house Puts one core message on top, the pillars that hold it up and the proof under each, in the words buyers already use. ## Purpose - One job: build the messaging house for one company and write it to `## Messaging house` in `company-profile.md`.- Stage: make. It ends with its own check (this skill's house checks, then `voice-check` and `claim-check`), at most 2 rounds.- The house is a source for copy, not copy. It holds the claims every channel draws from; headlines and ads are written later from it.- It never invents proof, a quote or a customer. A pillar with nothing under it is marked `proof needed` and stays visible. ## Use when / not | Situation | Use || --- | --- || Positioning is set and no messaging house exists | this skill || Pages and ads say different things about the company | this skill, then rewrite from the house || Positioning changed (a `positioning` run returned `REVISE`) | this skill, to rebuild || No positioning yet | `positioning` first || A page or ad exists and you want to know if buyers get it | `message-test` || Turning the house into ad angles per channel | `ad-themes`, after this skill | ## Inputs | Input | Required | Ways to provide, in order | If missing || --- | --- | --- | --- || Positioning: alternatives, unique attributes, value, best fit, category | yes | `## Positioning` in the profile; the `positioning` handoff; the reader gives the 5 fields now | `BLOCKED`, routed to `positioning` || Buyer words: interview and call transcripts, reviews (own and competitors'), sales emails from buyers, survey free text | no, but the house is weaker without it | a connected call recorder or CRM notes; review exports; uploaded files; pasted text | the house is in our words only: `DRAFT`, and `message-test` before use || Buying group and moments | no | `## Buyers` and `## Buying moments` in the profile | pillar order by role is skipped; coverage `partial` || Approved claims | no | `## Claims and proof` in the profile | pillars can rest only on product facts the positioning names; every other pillar is `proof needed` until the reader gives a source | ## Profile read From `company-profile.md`: `## Company` (name, category), `## Product`, `## Buyers` (with `### Buying group`), `## Buying moments`, `## Positioning`, `## Messaging house` (the current one, if any), `## Voice`, `## Claims and proof` (with `### Never say`), `## Competitors` (mention policy), and every `## Learnings` entry for `messaging-house` with status `noted`, `fact` or `accepted`. No profile: ask in one message for the 5 positioning fields, the claims the company can prove with their sources, and any interview or review text. Say that `company-profile` and `positioning` save this next time. ## Steps 1. **Apply learnings.** Accepted learnings for this skill override the defaults. Name each one that fires. 2. **Check the foundation.** Read `## Positioning`. All 5 fields must be filled. If any is missing or `not available`, stop at `BLOCKED` and route to `positioning`, unless the reader gives the missing fields now (then mark coverage `partial: positioning from the reader, not from evidence`). A homepage alone is not positioning: it is the copy the house is meant to fix. 3. **Mine the buyer's words.** Read every source line by line. Copy phrases verbatim, 2 to 12 words; never summarise. Tag each phrase:    | Tag | Means | Made-up example |   | --- | --- | --- |   | `before` | The situation or pain before buying | "every contract waits a week for a lawyer" |   | `after` | The outcome they wanted or got | "first draft back the same day" |   | `worry` | An objection, risk or fear about switching | "lawyers won't change how they redline" |   | `trigger` | What made them start looking | "the deal desk doubled its volume" |   | `alternative` | What they used or compared | "the review portal nobody opens" |   | `value words` | How they describe the benefit | "redline in Word" |    Record the source id and the role where known. Source order, best first, and what to take from each: `references/mining-guide.md`. Then count which phrases repeat across sources:    ```   python3 scripts/phrase_bank.py --sources i1.txt i2.txt r1.txt r2.txt --min-sources 2 > bank.json   ```    Without Python, list each phrase with the ids of the sources it appears in and keep the ones in 2 or more. Say which way you ran it. A phrase is buyer language when 2 or more distinct sources use it: 1 source is one person's style. 4. **Lay the foundation.** Write the positioning statement from the profile fields: "For <best fit> who <problem>, <company> is the <category> that <value>. Unlike <main alternative>, it <lead attribute>." Use the `positioning` handoff's `statement` if it exists. The statement is the test every line above it must pass; it is not copy. 5. **Write the core message.** Write 3 candidates, then score each on 5 rules:    | Rule | Passes when | Why |   | --- | --- | --- |   | Short | 15 words or fewer | It has to fit in one breath and in a headline slot |   | Result and who | It names the result and who gets it | Buyers buy the result, not the mechanism |   | Buyer words | The result is said with a mined phrase (or its exact key terms) | Buyers recognise their own words faster than ours |   | Only us | It rests on the lead attribute or the value theme, so the main alternative couldn't say it | A line any competitor could use does no work |   | Clean | No empty word and no claim missing from `## Claims and proof` | It must survive both checks unchanged |    Pick the candidate that passes the most rules; on a tie, the shorter one. If a candidate needs a comma and a second idea, the second idea is a pillar. 6. **Raise the pillars (3 or 4).** Start from the doubts buyers raise: the `worry` phrases and the buying group's objections. Most B2B houses answer these:    | Doubt | Pillar answers |   | --- | --- |   | Will it work? | The outcome, with a measured result |   | Will it fit how we work? | Effort, risk, the stack, the team's habits |   | Is it worth it? | Cost against the return or the cost of staying put |   | Why this, over what we use now? | The lead attribute against the main alternative |    Rules:   - Each pillar answers one doubt; no 2 pillars answer the same one.   - Use a 4th pillar only when a 4th doubt shows up in 2 or more sources. 3 is the default: every pillar needs its own proof, and each extra one splits a page's or an ad's space further.   - Each pillar traces to a positioning field (a unique attribute or a value theme). A pillar that traces to nothing is a feature list in disguise: cut it.   - Each pillar has a name (2 to 4 words), a claim (1 sentence, 20 words or fewer), proof from `## Claims and proof` or a product fact with its source, the doubt it answers, the role it matters most to, and the mined phrases it uses.   - No proof: write `proof needed`, with the kind of proof that would fill it. Never fill it with an adjective.   - Never use a table-stakes attribute as a pillar. `positioning` lists them. 7. **Map roles and moments.** For each buying group role, the pillar to lead with and why (from their objection). For each high-priority buying moment, the pillar to open with and the phrase to open on. Skip with `not available` when the profile lacks the sections. 8. **Check.** Run the house checks, then `voice-check` and `claim-check` (see Loop). 9. **Write the section.** Replace `## Messaging house` in the template's format:    ```   Core message: <one sentence>    - Pillar: <name> | <the claim> | Proof: <proof, with source>   ```    If a house already exists and the reader confirmed it, show old and new side by side first (named stop). If the agent can't edit the profile file, give the section in one code block ready to paste and set `profile_section_written: false (ready to paste)`. ## Check and verdict | Verdict | Condition (first match wins) || --- | --- || `BLOCKED` | No positioning in the profile and the reader can't give the 5 fields; or no product facts || `DRAFT` | Any pillar is `proof needed`; or no buyer sources were mined; or a check still fails after 2 rounds || `READY` | Every house check passes, `voice-check` has no critical finding and `claim-check` has no critical finding | Named stops: - `BLOCKED`: stop and route to `positioning`, or ask for the 5 fields in one message.- Replacing a messaging house the reader confirmed: show old and new with what changed and why; write after the reader says yes. `DRAFT` writes the section with `proof needed` visible on each pillar that lacks proof. `READY` writes without a stop. ## Loop 1. **House checks**, with `python3 scripts/phrase_bank.py --house house.md --bank bank.json`, or by hand:    | Check | Fails when |   | --- | --- |   | Core length | Over 15 words |   | Pillar count | Fewer than 3 or more than 4 |   | Proof | A pillar's proof is empty, `not available` or `proof needed` |   | Empty words | Any word from the list in `references/mining-guide.md` |   | Buyer words | The core message or a pillar uses no mined phrase (skip when nothing was mined, and say so) |   | Distinct doubts | 2 pillars answer the same doubt |   | Traces to positioning | A pillar rests on no unique attribute or value theme |   | Names the alternative | The statement doesn't name the main alternative and the difference | 2. **`voice-check`** on the core message and pillar claims. If it isn't installed, check inline: no dashes, every phrase under the profile's words to avoid, sentences over 30 words, and the 3 copy tests (can I picture it, can it be proved wrong, could no one else say it). Fragments are fine in pillar names. 3. **`claim-check`** on every claim and proof. If it isn't installed, check inline: every number and comparison matches a line in `## Claims and proof` (same number, same thing, same period); anything in `### Never say` is cut; a claim with no source is cut or turned into the mechanism ("reads your CRM stages, never rewrites them" instead of "3x faster"). Fix what failed and check again. At most 2 rounds. Then hand over with what is still open, labelled. ## Learn Append to `## Learnings`: - the reader rewrites the core message or a pillar: a `proposal` with their reason; if the reason is a word the buyers use ("buyers say redline, not review"), also suggest adding it to `### Words to use`;- results come back for a pillar, such as `message-test` scores, a buyer panel, or ad and page results by pillar from `ad-themes` or `landing-page-critic`: a `fact` with its source;- a proof is added or expires: a `fact`, and update the pillar. Example: `- 2026-10-24 | messaging-house | fact | The "fit" pillar scored lowest on value in 2 panel tests | Why: buyer panel results, 2026-10 | Status: noted` ## Output ```Verdict: <READY | DRAFT | BLOCKED> · Coverage: <complete | partial (what's missing) | blocked> · Mining: <script | by hand | not run: no sources>Sources mined: <n> (<ids>) · Phrases kept: <n> in 2 or more sources Phrase bank (top 10)| Phrase | Tag | Sources | Role | Foundation: For <best fit> who <problem>, <company> is the <category> that <value>. Unlike <main alternative>, it <lead attribute>. Core message candidates| Candidate | Short | Result and who | Buyer words | Only us | Clean | Pick | The houseCore message: <line>| Pillar | Claim | Proof (source) | Doubt it answers | Positioning field | Role | Buyer phrases | Lead pillar by role: <role: pillar, why>Opening pillar by moment: <moment: pillar, opening phrase> Checks: house <n>/8 · voice-check <verdict> · claim-check <verdict> · round <1 or 2>Open: <pillars with proof needed, failed checks> ## Messaging house   (the section as written, or "not written: <stop>")Learnings applied: <list, or none>``` Then the handoff block. ## Hand off ```handoffskill: messaging-houseversion: 1.0.0verdict: DRAFTcoverage: partial (4 sources mined; no competitor reviews)for: <company>outputs:  core_message: Redline in Word, with your clause library beside you.  pillars: ["Works inside Word", "Your clauses, every time", "Faster first drafts"]  proof_needed: ["Faster first drafts"]  lead_pillar_by_role: {General counsel: Faster first drafts, Contract manager: Works inside Word}  profile_section_written: truenext: [message-test, ad-themes, landing-page-critic]open: [review times before and after, from 2 or more customers]``` Made-up values for illustration. `message-test` scores pages and ads against the core message and pillars; `ad-themes` turns each pillar and its proof into angles per channel. ## Failure modes | Problem | Do this || --- | --- || The reader wants a pillar for every feature | A pillar answers a doubt. Group features under the doubt they answer; list the rest as supporting points || The best phrase comes from a competitor's review | Use the buyer's words about the problem; never use a competitor's name or product terms || Buyers' words are crude or off-voice | Keep the meaning and the key terms; `voice-check` decides the final wording || A customer quote is the only proof | Use it only if `## Claims and proof` lists it as cleared, or the reader confirms it || The house contradicts `### Never say` | Cut the line; `claim-check` marks it forbidden || Every pillar is `proof needed` | The positioning has no proof under it. Verdict `DRAFT`; suggest rerunning `positioning` with deal evidence | ## Influenced by - Joanna Wiebe's message mining (Copyhackers): reading reviews, calls and comments line by line, highlighting the phrases buyers repeat, and tagging them before choosing what to say: https://wistia.com/blog/joanna-wiebe-conversion-copywriting- SHAKEUP's messaging house, the same structure as the free messaging house builder on our site: positioning as the foundation, one core message, pillars with proof.- What we changed: phrases count as buyer language only when 2 or more sources use them, pillars are chosen by the buyer doubt they answer and must trace to a positioning field, the core message is picked by 5 scored rules, and the house is checked by `voice-check` and `claim-check` before it reaches the profile.

How it thinks

Inputs in, a verdict out, and a stop only where you decide.

What we built

We pick pillars by the doubt each one answers, not by feature, and every pillar must trace to your positioning. A phrase only counts as buyer language when 2 or more sources use it, and the house passes the voice and claim checks before it reaches your profile.

It reads your company profile if you have one, or asks for only what it needs. Then it works through its steps, checks its own work and ends in a verdict.

The steps

  1. 01

    Read your learnings first

    Rewrites you made to past houses and results from pre-tests or panels change the defaults before it starts.

  2. 02

    Check the foundation

    The house stands on your positioning. Without it, the skill stops and sends you to positioning, because a homepage is the copy the house should fix.

  3. 03

    Mine your buyers' words

    It reads interviews, calls and reviews line by line, copies phrases exactly and tags each one. A phrase counts as buyer language when 2 or more sources use it.

  4. 04

    Pick the core message

    It writes 3 candidates and scores each on 5 rules. 15 words or fewer, the result and who gets it, buyer words, a line only you could say, and nothing either check would cut.

  5. 05

    Raise the pillars

    Each pillar answers a different doubt buyers raise, traces to your positioning and carries proof from your approved claims. A pillar with no proof is marked, never padded.

  6. 06

    Check every line

    The voice check and the claim check run on the core message and every pillar. What fails is fixed and checked again, at most twice.

  7. 07

    Write it to your profile

    The house goes into your profile, where ad themes, landing page reviews and the message pre-test read it.

What you need

  • Required

    Your positioning (alternatives, unique attributes, value, best fit and category)

    The positioning section of your company profile → The handoff from the positioning skill → The 5 fields, given in the chat

  • Optional

    Your buyers' words, from interviews, calls, reviews and emails

    A connected call recorder or CRM notes → Review exports or files you upload → Text you paste

  • Optional

    Your buying group and buying moments

    The buyers and buying moments sections of your profile

  • Optional

    Your approved claims

    The claims and proof section of your profile

What it can return

  • READY

    Every house check passes and neither the voice check nor the claim check finds a critical problem.

  • DRAFT

    A pillar is still "proof needed", no buyer sources were mined, or a check still fails after 2 rounds.

  • BLOCKED

    There's no positioning to build on and you can't give the 5 fields.

Where it stops for you

  • When there's no positioning to build on
  • Before it replaces a messaging house you confirmed

Example output

What you get back.

A run for a made-up company. Your output depends on your inputs.

Here's a run for a made-up B2B SaaS company that sells contract review software to in-house legal teams. It mined 4 interviews and 3 reviews.

Verdict

DRAFT

Coverage: partial (no competitor reviews mined)

PhraseTagSources
"redline in Word"value wordsI1, I3, R2
"clause library"afterI1, I4, R3
"another portal nobody opens"worryI2, R1
"first draft back the same day"afterI4, R2
PillarClaimProofDoubt it answers
Works inside WordLawyers redline in the document they already use, with nothing new to openWord add-in, product docsWill lawyers use it?
Your clauses, every timeApproved fallback clauses show beside each clause as you reviewClause library, product docsWill it catch our clauses?
Faster first draftsThe first draft goes back the same dayProof needed: review times before and after, from customersIs it worth it?
Foundation

for in-house legal teams who review contracts in Word, the company is the contract review software that puts the clause library beside the document. Unlike review portals, it needs no new login.

Core message

Redline in Word, with your clause library beside you.

Checks

house checks 7 of 8 (1 pillar needs proof), voice check passed, claim check passed with 1 claim held back until it has a source.

The handoff block the next skill reads
skill: messaging-house
version: 1.0.0
verdict: DRAFT
coverage: partial (no competitor reviews mined)
for: contract review company
outputs:
  core_message: Redline in Word, with your clause library beside you.
  pillars: ["Works inside Word", "Your clauses, every time", "Faster first drafts"]
  proof_needed: [Faster first drafts]
  profile_section_written: true
next: [message-test, ad-themes, landing-page-critic]
open: [review times before and after, from 2 or more customers]

What good looks like

  • A core message of 15 words or fewer, in your buyers' words
  • Pillars that each answer a different buyer doubt
  • Proof under every pillar, or "proof needed" in plain sight
  • Phrases listed with the sources they came from
  • No word from your list of words to avoid

Install

Use it in the tool you already have.

Pick your tool. Each tab has the shortest path we know of.

  1. 01
  2. 02
    In Claude, open Customize, then Skills.
  3. 03
    Choose +, then Upload a skill, and pick the ZIP. Code execution must be on in your settings.
  4. 04
    Ask for it by name, or let Claude use it when a task matches.

Questions

Before you add it.

What's a messaging house?

A single page that holds what your company says. One core message sits on top, 3 or 4 pillars hold it up, and proof sits under each pillar. Every page and ad draws its words from it.

Why 3 or 4 pillars?

Each pillar answers its own buyer doubt and needs its own proof. A 4th only goes in when buyers raise a 4th doubt in 2 or more sources.

What if we don't have proof for a pillar?

It stays in the house marked "proof needed", with the kind of source that would fill it. The skill never fills the gap with an adjective or a made-up figure.

Can we use it without interviews?

Yes, but the house will be in your words, not your buyers'. It's marked as a draft, and the message pre-test should check it before you use it.