buyer-read
Skill1 starsChanged 17 days agoMITView on GitHub
Use this skill when a finished page, deck, or piece of client copy is about to go to the client or be published, and a real ICP or buyer research document exists. Produces a ranked list of what would make the target buyer stop trusting it, read in character, with every finding citing a line in a real source, plus what to cut and a score. Catches what craft reviews miss: scenarios that can't happen in the buyer's week, an offer never plainly stated, missing price, unused proof, seller vocabulary. Also use for "/buyer-read", "review this as the buyer", "have the ICP look at this", "what would a CFO think of this page", "be hypercritical about this copy". Not for grammar or voice checks, and never for rewriting.
In this repository67 other files it gives its agents.
- AGENTS.mdAGENTS.md
- CLAUDE.mdCLAUDE.md
- morning-paper-outreachSkill
- client-activation-checkinSkill
- client-churn-autopsySkill
- client-win-backSkill
- late-payment-chaseSkill
- network-path-mappingSkill
- rate-increase-planSkill
- sales-call-debriefSkill
- weekly-funnel-reviewSkill
- punch-upSkill
- call-topic-miningSkill
- ceo-exec-assistantSkill
- og-image-from-templateSkill
- weekly-site-and-ad-reportSkill
- icp-reviewSkill
- weekly-content-ideas-from-callsSkill
- client-performance-reviewSkill
- cold-email-sequenceSkill
- email-content-reviewerSkill
- scoring-analysis-and-distributionSkill
- delivery-audit-discovery-callSkill
- directionSkill
- rundownSkill
- source-checkSkill
- work-reviewSkill
- demandbird-exact-time-linkedin-postSkill
- kondo-owed-reply-sweepSkill
- linkedin-export-contact-importSkill
- mail-messages-contact-touchesSkill
- riverside-short-social-scheduleSkill
- youtube-transcript-panel-pullSkill
- email-voice-setupSkill
- upsell-opportunity-briefSkill
- ad-creative-content-strategySkill
- klaviyo-account-auditSkill
- facilitator-notesSkill
- local-prospectingSkill
- weekly-linkedin-postsSkill
- linkedin-watchlist-briefSkill
- pipeline-by-scriptSkill
- pre-session-flash-cardSkill
- write-in-the-owners-voiceSkill
- reddit-gtm-sprintSkill
- process-captureSkill
- sales-call-market-analyzerSkill
- linkedin-shoutout-postSkill
- plain-language-copy-checkSkill
- revenue-goal-funnel-mathSkill
- case-study-buildSkill
- cash-runway-forecastSkill
- client-profitability-readSkill
- client-proposalSkill
- content-briefSkill
- quarter-goal-filterSkill
- win-to-anglesSkill
- activity-incentive-sprintSkill
- candidate-interview-coachSkill
- client-list-stack-pitchSkill
- The first 60 shown.
The file
We split this skill into the rules it gives an agent
10 rules. About 0 are common, 9 are distinctive, and 1 is very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈9
Distinctive · estimate
This author's own choices. Worth a look.
1
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 26
No ICP or buyer research? Stop.
- Very specific: line 29
Run the nine checks in references/persona-and-checks.md, one at a time.
- Distinctive: line 45
MUST refuse to run without a real ICP or buyer research document.
- Distinctive: line 46
MUST cite a real source for every finding, or cut it.
- Distinctive: line 47
MUST write "not in the documents" for any persona element the sources don't cover.
- Distinctive: line 48
MUST rank by cost, not by discovery order.
- Distinctive: line 49
NEVER rewrite the copy. This critiques.
- Distinctive: line 50
NEVER pad to reach a number of findings.
- Distinctive: line 51
NEVER present a failing read as finished.
- Distinctive: line 52
NEVER quote the worked example as evidence.
≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
Applies to finished copy, run last, before it goes to the client or goes live. Produces a ranked, fully sourced buyer read with a score, or a refusal when no real buyer document exists.
Craft reviews grade what is on the page. They can't see what is missing, and they can't tell whether the writer understands how the buyer's business runs. This puts on the buyer's body, reads the piece cold, and reports where they would have stopped believing it.
The play
- Gather the piece (file, URL, or pasted text; if a URL won't open, ask for the text), the ICP or buyer document, any real buyer quotes (call transcripts, sales notes, reviews, survey answers), and client context (profile, proof, competitors, brand voice). Read references/worked-example.md before the first pass.
- No ICP or buyer research? Stop. Say what's missing and ask for it. Don't proceed on assumptions, even if the user asks you to guess.
- Build the persona before reading a word of the copy, following references/persona-and-checks.md. Where real buyer quotes exist they outrank the ICP: build from what buyers said, use the ICP for gaps, and report any disagreement between the two.
- Cold walk. Read the piece top to bottom in first person, in character, without looking at any brief or rubric. Note where you'd stop, disbelieve, get confused, or feel talked down to. Fix nothing yet.
- Run the nine checks in references/persona-and-checks.md, one at a time. Most produce nothing on most pieces, and that's correct. A check missing its source is marked "not run" with the source named.
- Evidence pass, separately from step 5. Attach a source to every candidate finding: a quoted line from the buyer document, a moment in a transcript, a documented competitor fact. A failed cold question needs no buyer line; cite the piece and what it never answers. Anything else with no citation gets cut, not softened. Then drop anything that is a writing preference wearing a persona.
- Rank by cost: the finding that loses the most buyers goes first. Own the writer's errors in a sentence. Name which existing section pays for the additions, or say the piece is too short to have one. Findings that cost conversions go above findings that are survivable.
- Score in a separate pass with references/scoring-rubric.md, then revise the lowest dimensions and re-score the whole read. Maximum three cycles. The rubric says what a stuck dimension means.
- Deliver, in order: the persona in two or three sentences with citations; ranked findings, each with what breaks, the evidence line, and the fix or decision it forces; a second tier; what to cut; questions for the client; what was cut in the evidence pass, one line each; the score sheet and
Buyer read: __/100, weakest <dimension>, cycles run: N.
What good looks like
- First thing to check that most people skip: the three cold questions. What am I being sold, what does it cost, show me one you've built. A beautifully written page that can't answer all three has an absence problem.
- The second thing: domain reality. A scene that can't happen in the buyer's week (a dead phone on their busiest day) triggers the deadliest objection there is: these people don't understand my business.
- The mediocre version: a persona that reads "a busy owner," ten findings that are style notes in costume, ranked in the order they were noticed, with no citations and nothing named to cut. It looks rigorous and is unfalsifiable.
- The read is right when every finding survives the question "where does it say that?", the persona is named, sized and situated from the documents, and it scores 85/100 or better with evidence traceability at 8 or higher.
- A three-finding read that passes beats a padded ten. If the piece is mostly fine, say so.
Rules
- MUST refuse to run without a real ICP or buyer research document. An invented persona produces confident, plausible, unfalsifiable findings, the worst failure of this skill, and it looks exactly like success.
- MUST cite a real source for every finding, or cut it.
- MUST write "not in the documents" for any persona element the sources don't cover. Invent only the name.
- MUST rank by cost, not by discovery order.
- NEVER rewrite the copy. This critiques; it names the fix or decision and stops.
- NEVER pad to reach a number of findings.
- NEVER present a failing read as finished. Deliver it marked not done, with the score sheet and the one line blocking the pass.
- NEVER quote the worked example as evidence.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

