Check Strategy

---
name: check-strategy
description: Judge whether the piece's argument holds — right buyer at the right awareness and sophistication, sections answering the buyer's real questions in a real order, and nothing starring that shouldn't lead. A strategic judgment gate argued from the buyer's perspective, not a checklist. Findings block. TRIGGER on the assembled selection at the assembly step (plan-copy 1d) before any plan file is written, and again on the built piece in the final pass, or when the chief runs the strategy gate. DO NOT TRIGGER for the reader's emotional reaction or belief (buyer-review), the line-level joins between sections (check-structure), or the CTA path (cro-review).
---

# Check Strategy

One Process: judge whether the argument is built right, before anyone reacts to it. This owns the argument's CONSTRUCTION — is this the right buyer, read at the right stage, answered in the right order. buyer-review owns the reader's REACTION to the built argument; this owns whether the argument should have been built this way at all. It runs twice, on two different objects:

- **1d assembly pass** — judges the ASSEMBLED PACKET plan-copy 1d hands it, before any plan file is written: the `strategies/` set carrying ALL strategies (each with its problem/buyer/offer shape/positioning, intervention type, and grading), the picked strategy, the option picks fanned inside it, the buyer hypothesis, the awareness and finalized sophistication reads with their citations, each pick's rating and market measurements, and the three declared strategist judgments. plan-copy 1d owns the packet's contents; this gate judges what it carries. A packet carrying only the picked strategy, without the whole set, is rejected — the breadth check below cannot run without it.
- **Final pass** — judges the BUILT PIECE, section by section.

Findings block — a failed strategy gate stops the piece.

### Judge construction, not existence
This gate judges whether the argument is BUILT right; it never judges whether the market, buyer, problem, and language it rests on EXIST — that is check-reality's category, run by a separate reality-reviewer, which returns a 1-to-100 existence rating per item. This gate reads those existence ratings as inputs riding on each item; it never re-runs the attack and never validates a framing merely because its statements could theoretically happen. Judge construction on the landscape as rated.

This is a strategic JUDGMENT, not a checklist. Given the assembled selection — or the built piece — the buyer, and the evidence, judge from the buyer's perspective — never tick a list of rules. The gate's quality is the system's defense against taste failures: a sharp gate catches them here, so no downstream rule has to.

## 1. Right argument for this buyer

Is this the right argument for this buyer, at this awareness and sophistication? Judge whether the direction is aimed at the buyer the selection names, read at the stage that buyer is actually in, and making the argument that buyer needs — not the one that was easy to make. Weigh the opening move against how much the reader already knows and how tired the market is of the promise (the ladders in copywriting's awareness reference). Fail a direction that assumes the reader arrives pre-wanting the product's angle — people arrive curious and wanting their problem solved, and the piece must build the desire it closes on.

### Judge the chosen pairing against the landscape, on the numbers
The packet targets one problem-by-buyer pairing out of a rated landscape, not a single pre-chosen buyer. Judge whether that pairing is the right pick from the landscape — a higher-rated problem-by-buyer pairing the picks ignored is a failure, not a detail. Judge on the ratings, existence ratings, and market measurements the packet carries and the reasoning behind them, never on an agent's disposition or a bare true/false. A pick that leans on "this is the right one" without its grading fails.

### Gate the declared judgments as judgment, not as fact
At the 1d assembly pass the packet labels three calls as the strategist's judgment: the buyer clustering (which groups formed each hypothesis), the promise-class grouping (which competitor claims were counted as the same promise, setting sophistication), and the objection order. Judge each against the buyer — is this how the buyer's speech actually clusters, is this the same promise to this buyer, is this the order the buyer's questions arise — and never accept any as a fact from the evidence. A judgment that only reads clean because it was asserted as evidence-derived fails.

### Trace every behavioral assertion to one speaker chain
Every behavioral assertion in a strategy narrative — what the buyer tried, what failed, what they did next, what it cost — must trace to ONE named speaker chain in the evidence. An assertion no single speaker's chain carries is a composite stitched across speakers, and it fails as a blocking finding, however coherent the person it describes reads.

### Trace every headline and positioning word to buyer vocabulary
Every key noun and verb in a headline or positioning line must trace to quoted buyer vocabulary in the cited evidence. Strategist vocabulary — "diagnosis", "verdict", "audit" — fails unless a buyer said it. A metaphor fails, including time and space figures ("before ad money touches it"). A line that only reads sharp because the strategist supplied the words, not the buyer, fails the gate.

## 2. Right question sequence

Does the order answer the buyer's real question sequence? This owns whether the ARGUMENT's order sells — whether the sequence of points is the one that moves this buyer to buy. The questions must be the buyer's own, traced to problem research; a sequence that conveniently sells the product fails at either pass.

On the FINAL PASS, walk the built piece as the one held buyer, section by section, and judge whether each answers the question the buyer is actually carrying at that point — in the order the buyer asks them, not the order that is easy to teach. A section answering a question no buyer is asking, or an order that follows a teaching sequence, fails. The mechanical join between two sections — whether the seam is smooth — is check-structure's, not this gate's.

On the 1d ASSEMBLY PASS there are no sections yet; what the packet carries is the objection ORDER as the strategist's declared judgment, not an evidence-derived fact. Judge that order against the buyer — is this the order the buyer's real questions arise in — and never accept it as if the evidence set the sequence.

## 3. Right thing leading

Does anything star that shouldn't, or hide that should lead? Judge what the piece puts first and repeats against what this buyer came for. Argue it from the buyer's perspective — what they want to see, what would make them buy — never from a fixed rule about any one feature or section.

### Report a set that collapses to one intervention
Intervention type is defined by what the buyer RECEIVES or DOES, never by the label the strategist gave it: an explanation-of-cause and a list-of-changes are ONE type (the buyer is told about the page); a done-for-you build is another; a tool to do it themselves another; more traffic another; a checkout or payment fix another. The packet carries each strategy's classification; judge the set against these received-actions. A set whose strategies all reduce to one received-action — three labels ("diagnosis", "redesign", "education") on the same "told about the page" action — is collapsed, however distinct the labels read. Relabeling within one received-action is convergence, not divergence. A packet that omits a strategy's classification cannot be judged here until it carries one. Judge the whole set the packet carries, not the picked strategy alone, for its breadth: whether it spans at least 2 distinct problems, at least 2 distinct buyer groups, and distinct received-action classes. A thin candidate set — below any of these — is a BLOCKING FINDING reported to the chief, never a set this gate kills on its own. The gate reports the thinness with its reasoning; the chief rules — proceed knowingly on the set as-is, or re-commission for more breadth.

### Reject a folded pick at the assembly gate
At the assembly step (plan-copy 1d), the picked strategy must have its `strategies/` file and its own owner pick, and every option set inside it — Mechanism, big idea, any refined offer detail — must have its own option file and its own owner pick. A post-pick offer or positioning CHANGE is not an option change: plan-copy replaces the whole strategy with a new one and a new pick, and the packet must carry that newly picked strategy alone, never the original plus a patch. A selection where the picked strategy or any option lacks its file and pick, or where a post-pick offer or positioning change rides as a patch on the original strategy, fails the gate — coherent though it may read.
Never: passing a selection whose Mechanism, big idea, or refined offer detail was folded in without an option file, or whose post-pick offer or positioning change patches the original strategy instead of replacing it.

## 4. Findings block

Write every failure as a blocking finding in review-copy's finding format (check, severity, location, finding, suggested fix). Argue each from the buyer's perspective, so revise can answer it. A failed strategy gate stops the piece.

### Record the 1d gate in a findings stub
At the assembly step (plan-copy 1d), write the gate's findings record to `findings/StrategyGate.md` — the selection hash, the date, what was judged, and any blocking findings the gate raised (none, when the selection is clean). The stub is the gate's findings record, not a literal pass flag: it evidences that the gate ran and carries whatever it found. plan-copy's file-writing step (1e) requires this stub before any plan file is written, and plan files proceed when no blocking findings remain on the current selection — so a silent skip of the gate cannot look like a run.

Verification: the chosen problem-by-buyer pairing judged against the rated landscape and on the packet's ratings, existence ratings, and market measurements rather than disposition; the argument judged for the right buyer at the right awareness and sophistication, the section order judged against the buyer's real question sequence, and what leads and repeats judged against what the buyer came for — each argued from the buyer's perspective; every failure recorded as a blocking finding in review-copy's format.