Skip to content

The design language that makes your AI harness better at design.

JavaScript

Short comp prompts: carry the product, leave the look to the world (#975)

Paul Bakaus7bbaf5710c56
diff --git a/skill/agents/impeccable-asset-producer.md b/skill/agents/impeccable-asset-producer.md
index e1284f82b..93d0bbbe7 100644
--- a/skill/agents/impeccable-asset-producer.md
+++ b/skill/agents/impeccable-asset-producer.md
@@ -22,7 +22,7 @@ Do not redesign. Preserve the reference's visual role, silhouette, palette, ligh
## Decision Comps
-When the parent hands you a decision card packet instead of an approved mock, the job is one comp: one card, one file, written to the card's declared `comp` path the moment it renders. The parent runs several of you in parallel, one per card, so this card is your entire contract; generate first, plan never, because the file on disk is the deliverable and the decision page is waiting on it. Work from the card's structured fields and PRODUCT.md alone; report a card too thin to brief a comp, never pad it from imagination. Render the card's direction as a north-star comp at full fidelity: the requested surface's first viewport, prompt led by the surface's own structure (regions named in order with their scale relationships, never the world's atmosphere), fully committed in the card's own palette, type character, and material world. A native app or mobile-first surface is a portrait frame at its device viewport, never a landscape default. Every sibling renders at the same full fidelity in its own grammar, one surface, one aspect; equal commitment keeps the comparison honest. Real product name and real content only; never invent commercial claims, prices, benchmarks, or dates PRODUCT.md does not carry. Exclusions bind those claims, never a medium the card's own world has not excluded: a subject that lives in photographs keeps its photographs. Write the prompt sidecar beside the file. Return one line naming the path and any deviation, nothing more. Everything below this section is the asset-production job; none of it applies to a decision-comp run.
+When the parent hands you a decision card packet instead of an approved mock, the job is one comp: one card, one file, written to the card's declared `comp` path the moment it renders. The parent runs several of you in parallel, one per card, so this card is your entire contract; generate first, plan never, because the file on disk is the deliverable and the decision page is waiting on it. Work from the card's structured fields and PRODUCT.md alone; report a card too thin to brief a comp, never pad it from imagination. Render the card's direction as a north-star comp at full fidelity: the requested surface's first viewport, prompted in the short shape visualize.md gives (the product and the visitor's job with real content, what leads, the parts in reading order, the frame), with the look taken from the card's world: its card images attached as references when the packet names them, otherwise two to four sentences of concrete look. A native app or mobile-first surface is a portrait frame at its device viewport, never a landscape default. Every sibling renders at the same full fidelity in its own grammar, one surface, one aspect; equal commitment keeps the comparison honest. Real product name and real content only; never invent commercial claims, prices, benchmarks, or dates PRODUCT.md does not carry. Exclusions bind those claims, never a medium the card's own world has not excluded: a subject that lives in photographs keeps its photographs. Write the prompt sidecar beside the file. Return one line naming the path and any deviation, nothing more. Everything below this section is the asset-production job; none of it applies to a decision-comp run.
## Review handoff
diff --git a/skill/reference/mode-operate.md b/skill/reference/mode-operate.md
index adfe966ae..833775ba8 100644
--- a/skill/reference/mode-operate.md
+++ b/skill/reference/mode-operate.md
@@ -14,7 +14,7 @@ Write each direction's first viewport as the product's shell first: the navigati
The comp is the product screen, in the direction's committed palette: its shell and navigation with a distinct area for each job and, where there is data, filters over what they scope and charts and tables with clear states. The world enters only through type, palette, density, and the one signature move; it never supplies the page's organisation, so a working screen never reads as a newspaper, a poster, or any other artifact of the world. Every control is a standard one the visitor already knows how to use, a web component on the web and the platform's native control in an app: buttons, tabs, inputs, selects, filters, tables with sortable headers, navigation that reads as navigation. Never a costume of the world's own interface: a terminal's function-key bar or command prompt, an instrument's bezel, ruled ledger paper standing in for a table. A comp an engineer could not build from the platform's standard components is a picture of the tool, not the tool. The finish bar is the best product in the category: an aligned grid, consistent spacing, a real type scale, and clear hierarchy. The signature move adds to a screen that is already that clean; it never stands in for it.
-Prompt with the screen's real content: the product name, realistic data with plausible values and units, real labels, and one state worth seeing (an alert, a degraded service, an empty filter result), never placeholder tiles.
+Prompt with the screen's real content: the product name, realistic data with plausible values and units, real labels, and one state worth seeing (an alert, a degraded service, an empty filter result), never placeholder tiles. State colours (red for critical, amber for degraded) sit on top of the world's palette only where they carry meaning, and a headline number that moves over time shows its recent history, with the onset marked when something changed.
Comps two and three vary the product's organisation, never its world coverage: the navigation model (side rail, top bar, command-first), how the work areas divide the screen, which data leads, the density, and where the signature move sits. Each must be recognisably different from the first at thumbnail size; a variant only its prompt can tell apart adds no choice. A variant that adds world material, texture, or ornament goes further in the wrong direction; on a working screen, more world is a costume.
diff --git a/skill/reference/new-work.md b/skill/reference/new-work.md
index 6e24a4afb..985c0f5f6 100644
--- a/skill/reference/new-work.md
+++ b/skill/reference/new-work.md
@@ -38,19 +38,19 @@ Keep the visual system fixed. Derive five to seven materially different structur
`{{scripts_path}}/impeccable concept-seed --scope surface --mode <mode>`
-The script deals three of your structures; the dice pick which three reach the user, breaking the ranking rut while the user keeps a real choice. Present them on the decision page as full cards of equal salience, the dealt lead under kicker THE ROLL, with steer and re-roll; the user locks one. No canon card and no pick card at surface scope: the world is settled, so every card visualizes composition, not identity. With image generation and a comp-led default (`.impeccable/config.json`; the build-path paragraph below), each card declares a `comp` under `.impeccable/mocks/decision/`, generated after serving, in reading order, under [visualize.md](visualize.md)'s comp discipline and the MODE RULES block concept-seed printed for this surface. Anchor each comp on the established identity: pass a screenshot of a representative existing page as a reference image (the harness image tool's input image, or `impeccable generate-image --ref`) with a prompt that leads with the new surface's structure and names DESIGN.md's palette, type, and component character; prose paraphrases of a design system drift, pixel references do not. Without image generation, or under a code-led default, each card carries a `wireframe` schematic (`impeccable serve-question --schema`) the page draws itself. Locking a card is the approval and sets the build path: a locked comp builds comp-led with that comp as the approved comp, discharging [visualize.md](visualize.md)'s three-option round with no second approval point; a locked wireframe builds code-led, its ambition carried by the direction contract. Never run the script for a local extension or a precisely specified narrow request; shape those directly.
+The script deals three of your structures; the dice pick which three reach the user, breaking the ranking rut while the user keeps a real choice. Present them on the decision page as full cards of equal salience, the dealt lead under kicker THE ROLL, with steer and re-roll; the user locks one. No canon card and no pick card at surface scope: the world is settled, so every card visualizes composition, not identity. With image generation and a comp-led default (`.impeccable/config.json`; the build-path paragraph below), each card declares a `comp` under `.impeccable/mocks/decision/`, generated after serving, in reading order, under [visualize.md](visualize.md)'s comp discipline and the MODE RULES block concept-seed printed for this surface. Anchor each comp on the established identity: pass a screenshot of a representative existing page as a reference image (the harness image tool's input image, or `impeccable generate-image --ref`) with a short prompt in visualize.md's shape; prose paraphrases of a design system drift, pixel references do not. Without image generation, or under a code-led default, each card carries a `wireframe` schematic (`impeccable serve-question --schema`) the page draws itself. Locking a card is the approval and sets the build path: a locked comp builds comp-led with that comp as the approved comp, discharging [visualize.md](visualize.md)'s three-option round with no second approval point; a locked wireframe builds code-led, its ambition carried by the direction contract. Never run the script for a local extension or a precisely specified narrow request; shape those directly.
### Create or replace the visual world
1. Name the product's unique mechanism in one sentence, the audience's real scene, its cultural home, and what this first surface must prove. Note the page this category always ships and its predictable opposite; both are the rut, kept out of the seven-candidate list. A brief that paints its own picture, a product name, a titled artifact, a governing metaphor, adds its literal reading to the rut: spend at most one candidate on it and derive the rest from elsewhere in the audience's world.
-2. From that cultural world, list seven concrete visual systems, artifacts, places, or rituals the audience knows by heart, each with one line on why it resonates and can carry the mechanism, ordered by resonance. The audience's world includes its graphic and screen traditions, not only its physical objects: the notation, publications, identity programs, data graphics, and interfaces it reads daily. On an Operate or Read surface, the seven still come from anywhere in that world (its graphic traditions, places, rituals, and artifacts), never from the tools and instruments the audience operates: a terminal, a scope, or a control panel handed to a working screen comes back as a costume of that tool. A nameable abstract system (a school of poster, a documentation standard) is as concrete a candidate as any artifact. On a Persuade or Experience surface, the tools and instruments the audience operates count toward the rut: spend at most one candidate on them. Also ask what this thing would look like as a physical object and what its world looked like before the web. Near-duplicates count once. When more than three of the seven share one material family, the derivation stopped at the subject's most obvious artifact; dig until the list spans at least three families.
+2. From that cultural world, list seven concrete visual systems, artifacts, places, or rituals the audience knows by heart, each with one line on why it resonates and can carry the mechanism and one on how it looks (its ground and main colours, its type voices, the one signature that makes it recognisable), ordered by resonance. The audience's world includes its graphic and screen traditions, not only its physical objects: the notation, publications, identity programs, data graphics, and interfaces it reads daily. On an Operate or Read surface, the seven still come from anywhere in that world (its graphic traditions, places, rituals, and artifacts), never from the tools and instruments the audience operates: a terminal, a scope, or a control panel handed to a working screen comes back as a costume of that tool. A nameable abstract system (a school of poster, a documentation standard) is as concrete a candidate as any artifact. On a Persuade or Experience surface, the tools and instruments the audience operates count toward the rut: spend at most one candidate on them. Also ask what this thing would look like as a physical object and what its world looked like before the web. Near-duplicates count once. When more than three of the seven share one material family, the derivation stopped at the subject's most obvious artifact; dig until the list spans at least three families.
3. Turn that material into complete directions: each joins a reusable visual world to a concrete first-surface experience.
4. Run `{{scripts_path}}/impeccable concept-seed --scope direction --mode <mode>` and follow what it prints. No substitute, no skip: on a new or replacement world, writing artifact code before this script has run and its assignment is acknowledged is a contract violation, whatever the harness, the model, or the time pressure; the roll is what keeps every run from converging on the category default. The script assigns the direction to build and deals catalog challengers. Fuse each challenger before judging it: the challenger supplies the form and its system grammar, the product supplies every fact, clarity wins conflicts. Weigh fused challengers against the assigned direction on exactly two axes, audience identification and product clarity. Losing to strong grounded material is a valid outcome; beating a thin or tool-monoculture list is the point. Close with a verdict per challenger, decided before any borrowing: wins (beats the assigned direction on both axes; becomes the build candidate), competitive (holds one axis; stays a full alternate), or declined (loses both). A declined challenger is not spent: name the one discipline of its system the assigned direction lacks, and raise the assigned direction to match before presenting it. A donation transfers ambition and system discipline (a palette's total commitment, a grid's density courage, a form's structural honesty), never the challenger's clothes. A component treatment, such as a button's shadow or a display face, is clothes, not discipline; a lifted motif is a costume note, not a raise, and one world owns the page. Write each raise into the presented direction as its own line, named for its donor; a raise nobody can read did not happen. <!-- rule:skill-concept-procedure --> <!-- rule:skill-verdict-and-donation -->
5. Present one direction, fully committed and already raised by the hand it beat, raises visible as named lines: world, first viewport, visitor path, signature interaction, cross-surface reach, honest risk. Route each challenger by verdict: winning and competitive challengers are full alternates with their QUALITY BAR cards and one-line case; declined challengers render demoted, compact and quiet, each carrying its verdict and what the direction kept from it, never full-size, never silently dropped, still adoptable on request. The verdict informs the user's choice, never pre-empts it; the demoted row is the hand's proof of judgment. A hand holds at most three full-card challengers: when the roll deals more, the three strongest join and the rest wait in the re-roll pool, noted in one line; dropping a challenger from the hand itself takes a named product-truth failure, disclosed. Add one card for your own top-ranked grounded candidate when it is not the assigned direction, kicker IMPECCABLE’S PICK, same anatomy as every card, with an honest risk line naming its familiarity when true: the strongest grounded direction is often where most runs in this category land, and the user deciding that trade is the point of showing it. Familiar and effective is a legitimate destination, not a failure of nerve; the pick card and the standing exit serve it at two depths. One pick card, never two, never a ranked list: a lineup of your candidates hands selection back to a taste function and invites the safest card. The pick never takes the lead position; when the dice assign your top candidate there is no pick card, and the assigned card notes it topped your list. Add re-roll with an optional one-line steer, in three registers: plain (a fresh hand, same spread), safer (your remaining conventional grounded candidates plus the canon against named competitors), bolder (foreign forms only, at full commitment). The register is the user's steering on the familiar-to-bold axis, never yours to pre-select; when the answer carries one, re-run the seed with `--register <value>` and the next `--reroll` round, and follow what it prints. A user saying "bolder" or "safer" while a direction round is open means these registers, never the bolder or harden commands. The two channels share this structure and differ only in richness: cards and boards on the decision page, names and one-liners through the structured tool, whose option list carries the assigned direction, the pick, the winning and competitive challengers, and the standing exit last; declined challengers fold into the assigned option's description as their kept lines, so the raise survives the text channel. <!-- rule:skill-pick-card-one-only -->
The standing exit: every direction round offers one quiet, permanent alternative, the category standard, played straight. It is the user's door, never yours: never recommend it, never weigh it against the roll, never let it soften the dealt directions; the counterweights bind the unchosen default, not the chosen one. When the user takes it (the canon action, a safer-steer, or plain words asking for the familiar or competitor-like path), convention becomes the commitment: ask once for two or three products this should sit alongside, make their craft level the bar, and execute the canon at full fidelity, without irony or smuggled quirk. Record a standing preference as a brand commitment in PRODUCT.md. <!-- rule:skill-canon-standing-exit --> Re-roll eliminates every direction already shown, grounded and challenger alike; after two consecutive re-rolls, ask what quality is missing. Re-roll on your own only on named factual grounds, when the assigned direction cannot carry the product's truth or task; taste is never grounds. The user may re-roll freely, and a user- or brief-pinned direction beats the roll, always. Resolve collisions field by field: preserve every user- or brief-pinned constraint. In dimensions the brief leaves open, the assignment still binds through its topology, controls, state vocabulary, and ritual; when only its materials conflict with a pinned visual direction or PRODUCT.md brand commitment, translate that material expression and name the translation in the presented direction. A look mismatch is not grounds to re-roll. <!-- rule:skill-assigned-plus-reroll --> Present the decision visually: write an options payload with the assigned direction leading, its raised lines included; the pick card when one exists; the dealt challengers as alternates with their QUALITY BAR cards, verdicts, and kept lines; re-roll with its safer and bolder registers; steer; canon enabled; and `buildPath` carrying the recorded default with `toggle: true` whenever image generation exists (details in the build-path paragraph below). A degraded roll with no challengers still uses the page, as a single text-only card with re-roll. Give every card the same anatomy: thesis, palette, materials, first viewport, honest risk, and the challengers' case lines (`--schema` prints the exact shape); the page renders identity from these fields, demotes declined challengers to their row on its own, and a challenger's catalog image rides as labeled inspiration, never the promise of the build. Author `canonCard` too: the category standard as one honest card, same anatomy; the page keeps it subordinate, and the counterweights still bind you. Run `{{scripts_path}}/impeccable serve-question --start --payload <file>` (`--schema` first for the payload shape). It daemonizes, prints the page URL and a key, and exits; open that URL for the user, in-app browser first, then the system opener, then showing the URL. Collect the choice with `--wait --key <key>`, repeating while it exits 3; the ANSWER prints as JSON. An ANSWER of `{"optionId":"reroll"}` keeps the server alive and the page open on a loading hand: rerun concept-seed with the same `--scope` and `--mode` plus `--from <seed-key> --reroll <n>` (1 on the first re-roll, counting up), build the next payload, deliver it with `--update --key <same key> --payload <file>`, then return to `--wait` on that key. Never `--start` a second server or fall back to chat here: either strands the open tab on a hand that never arrives. Exit 4 means the page closed unanswered: re-present once through the structured question tool, and with no answer there either, proceed unattended with the assigned direction and state the assumptions. A harness that can leave a shell blocked in the background may run the script without `--start` and let it auto-open and block. A harness that cannot hold a blocking `--wait` (it detaches long-running commands without ever returning their output, or reaps a backgrounded child) takes the structured tool without starting the page and discloses the substitution in the first reply, so the tab is never left unanswered. A shell that hands back a still-running command as a session you can poll holds the wait: poll that session until it exits, and rerun `--wait` only after it exits without an answer. Never predict any other fallback: run the script, and only exit code 2 from starting it, or that wait check before starting, routes the decision to the structured tool; those are the fallback, never an error to retry. concept-seed closes every roll with this procedure, condensed, as a PRESENTATION block that names the recorded build path. <!-- rule:skill-visual-decision-page -->
-When image generation exists, every card also declares a `comp` path under `.impeccable/mocks/decision/`, the canon card included. Where the harness sandboxes its shell, start the page through the least-sandboxed command path it offers: a sandboxed shell cannot bind the board's port, and the first-attempt failure costs a retry every session. Serve the page first, then produce the comps; the page shimmer-waits per slot and the user may answer before they land. Load [visualize.md](visualize.md) before writing the first decision comp prompt, not at the comp round: each card's image is that direction's north-star comp at full fidelity under its comp discipline and the MODE RULES block concept-seed printed, and a prompt written before reading them misses every rule there: the requested surface as the real page, its first viewport and the start of the section below it, held like every comp to visualize.md's rule that a template skeleton stays the template whatever world paints it (your own pick included: a safe layout under a world skin is the template), structure-led prompt, real product name and real content, no invented commercial claims, in that card's own palette, type character, and material world, committed all the way. Once a round declares comps it has not generated yet, `impeccable serve-question --start` (and `--update`, and a flip to comp) prints a `NEXT read <path>/visualize.md` line; read that file then, before the first prompt. `--wait` prints `COMP SIDECAR MISSING` for any landed decision comp with no prompt sidecar (the image's full file name plus `.json`: `a.png` gets `a.png.json`); write the missing sidecar before you continue. `COMP STALE` names a file an earlier round left at one of this hand's slots; regenerate it there. Generation takes the same time at any fidelity, so an unfinished draft pays comp cost for draft quality; fairness between cards is equal fidelity in each card's own grammar, one surface, one aspect, never shared unfinishedness. The frame's aspect is the surface's own: portrait at device viewport for a native app or mobile-first surface, landscape for desktop web; the decision page adapts to either, and a phone screen comped landscape is a broken frame, not a neutral default. Produce in reading order, the assigned card, then the pick, then the full-card hand, then canon, each file written with its prompt sidecar the moment it is done, so a re-roll's spend front-loads onto the cards read first; declined challengers get no comp, their catalog thumb is their face. With parallel subagents, fan out one agent per card: each spawn is the shipped asset producer with a single-comp packet, that card's fields, PRODUCT.md, the shared frame, the path to visualize.md for its comp discipline, the COMPS section of the MODE RULES block copied verbatim (the subagent never sees the roll's output), and the card's declared path, up to four in flight. Regenerate inline any slot still empty when its agent returns; drop without ceremony any slot still empty when the user answers. No other supervision is owed. Without parallel subagents, generate in the main thread after serving, same order, and let the harness's own generation display carry the progress; the wait for the answer follows the last file. The chosen card's comp is not spent by the choice: comp-led, it enters the comp round as compositional option one; code-led, it returns at the finish review as the critique reference, what the image dared that the build did not. Unchosen comps stay in `.impeccable/mocks/decision/` as the round's spent hand; they carry no approval and imply none. With no image generation, cards carry their identity in palette chips and facts, and that page is complete, not a lesser version; the page then also demotes every challenger's catalog art to a labeled thumbnail on its own, because salience must encode the verdict, never the accident of which cards have images. <!-- rule:skill-decision-comps-full-fidelity --> <!-- rule:skill-salience-parity -->
+When image generation exists, every card also declares a `comp` path under `.impeccable/mocks/decision/`, the canon card included. Where the harness sandboxes its shell, start the page through the least-sandboxed command path it offers: a sandboxed shell cannot bind the board's port, and the first-attempt failure costs a retry every session. Serve the page first, then produce the comps; the page shimmer-waits per slot and the user may answer before they land. Load [visualize.md](visualize.md) before writing the first decision comp prompt, not at the comp round: each card's image is that direction's north-star comp at full fidelity under its comp discipline and the MODE RULES block concept-seed printed, and a prompt written before reading them misses every rule there: the requested surface as the real page, its first viewport and the start of the section below it, held like every comp to visualize.md's rule that a template skeleton stays the template whatever world paints it (your own pick included: a safe layout under a world skin is the template), a short prompt in visualize.md's shape, real product name and real content, no invented commercial claims, in that card's own palette, type character, and material world, committed all the way. Once a round declares comps it has not generated yet, `impeccable serve-question --start` (and `--update`, and a flip to comp) prints a `NEXT read <path>/visualize.md` line; read that file then, before the first prompt. `--wait` prints `COMP SIDECAR MISSING` for any landed decision comp with no prompt sidecar (the image's full file name plus `.json`: `a.png` gets `a.png.json`); write the missing sidecar before you continue. `COMP STALE` names a file an earlier round left at one of this hand's slots; regenerate it there. Generation takes the same time at any fidelity, so an unfinished draft pays comp cost for draft quality; fairness between cards is equal fidelity in each card's own grammar, one surface, one aspect, never shared unfinishedness. The frame's aspect is the surface's own: portrait at device viewport for a native app or mobile-first surface, landscape for desktop web; the decision page adapts to either, and a phone screen comped landscape is a broken frame, not a neutral default. Produce in reading order, the assigned card, then the pick, then the full-card hand, then canon, each file written with its prompt sidecar the moment it is done, so a re-roll's spend front-loads onto the cards read first; declined challengers get no comp, their catalog thumb is their face. With parallel subagents, fan out one agent per card: each spawn is the shipped asset producer with a single-comp packet, that card's fields, its world's card images when the NEXT line names them, PRODUCT.md, the shared frame, the path to visualize.md for its comp discipline, the COMPS section of the MODE RULES block copied verbatim (the subagent never sees the roll's output), and the card's declared path, up to four in flight. Regenerate inline any slot still empty when its agent returns; drop without ceremony any slot still empty when the user answers. No other supervision is owed. Without parallel subagents, generate in the main thread after serving, same order, and let the harness's own generation display carry the progress; the wait for the answer follows the last file. The chosen card's comp is not spent by the choice: comp-led, it enters the comp round as compositional option one; code-led, it returns at the finish review as the critique reference, what the image dared that the build did not. Unchosen comps stay in `.impeccable/mocks/decision/` as the round's spent hand; they carry no approval and imply none. With no image generation, cards carry their identity in palette chips and facts, and that page is complete, not a lesser version; the page then also demotes every challenger's catalog art to a labeled thumbnail on its own, because salience must encode the verdict, never the accident of which cards have images. <!-- rule:skill-decision-comps-full-fidelity --> <!-- rule:skill-salience-parity -->
The execution contract, comp-led or code-led, is a workflow preference, not a per-surface decision; no round asks it. The recorded default rides every round and the page's toggle handles the exception. Read the default from `.impeccable/config.json` (`buildPath`), the gitignored `.impeccable/config.local.json` winning where one machine differs from the team's committed value; with neither, comp-led is the default whenever image generation exists. Author every direction and surface payload with `buildPath: { "value": <default>, "toggle": true }`; the page renders a footer toggle with the trade stated beside it, and the ANSWER returns `buildPath` plus `buildPathFlipped`. A flipped value binds that session only and is never written back, with one exception, the only question this preference ever earns inside a round (init records it up front on projects that get the chance): when `buildPathFlipped` comes back true on a project that records no `buildPath` at all, ask once after the round closes whether to keep it as the standing default. Either answer writes `.impeccable/config.json`; the answer picks the value, never whether to record one. Yes writes the flipped value; "no, just this once" writes the value they flipped away from, the standing default they just confirmed by declining. Ask on the flip, never on the untouched default: a user who left the toggle alone told you nothing. A declined offer nothing writes down is an offer the next session makes again. When the user asks in words to change the standing default, update the file without asking. **Comp-led**: the chosen card's comp is law, generated before building when it does not yet exist, and the finish review audits the build against it; boldest composition on the table, fix rounds expected, and the comp is non-optional, no silent skipping. **Code-led**: no comp of this page and no apology for it; the QUALITY BAR boards still calibrate finish, and the ambition moves into the written contract, the FIRST VIEWPORT block plus a named signature interaction and motion grammar, which the finish reviewer audits in behavior; code-led is not a discount on commitment. A code-led round still declares each card's comp path as a flip reserve: when the user flips the toggle to comp mid-round, `--wait` returns once with BUILD PATH FLIPPED while the page shimmers the slots; generate each open card's comp into its declared path then, lead first, and wait again. The flip back is free, and a comp that already rendered rides at the finish review as the critique reference. Without image generation there is no toggle and no choice: code-led is the only path, stated in one line rather than asked. The structured-tool fallback has no toggle either: it builds on the recorded default and asks nothing about the build path. The old two-card execution-contract round is retired; `followup: true` remains the general mechanism for delivering any later round over the same table via `--update`. <!-- rule:skill-build-path-round -->
diff --git a/skill/reference/visualize.md b/skill/reference/visualize.md
index 26fa7b273..5387bf221 100644
--- a/skill/reference/visualize.md
+++ b/skill/reference/visualize.md
@@ -10,12 +10,13 @@ The comp round runs inside the build's phase state: `impeccable build-phase star
Render three distinct high-fidelity north-star comps of the requested surface, saved under `.impeccable/mocks/` so they survive the session. When the chosen card has a decision comp, it is option one: it stays at its path under `.impeccable/mocks/decision/`, where the gate counts it through `--decision-comp`, and is never regenerated; render the other two. Comp at the surface's own viewport: portrait at device size for a native app or mobile-first surface, desktop landscape otherwise; a phone screen comped landscape misstates the composition before anything is built against it. Comps are the build thread's own work, never delegated: the thread that writes the prompts holds the direction's full context and has seen every comp when the build starts. Open every image by its workspace-relative path; sandboxed viewers reject absolute paths, and everything under the project root has a relative one. Base the comps on real content and the surface concepts already developed with the user. On an established world, anchor every comp on the real identity: capture a screenshot of a representative existing page and pass it as a reference image (the harness image tool's input image, or `impeccable generate-image --ref`); the prompt leads with the new surface's structure while the reference carries palette, type, and component character, because DESIGN.md words alone drift where a pixel reference does not. Name what the reference contributes and what it must not: chrome, palette, type, and component character carry over; the reference page's own content does not, and a banner, hero, or card lifted verbatim is the reference leaking, not fidelity. Three is the number: one comp invites rubber-stamping; the spread between three surfaces the composition worth building. The chosen card's decision comp is the first of the three: it already renders this direction at full fidelity under this discipline, so generate two more that vary what the first held fixed, from the direction in words rather than with the first comp as an input image (a reference image anchors the layout, and three comps collapse into one composition). How comps two and three vary from the first is the mode's call, in its COMPS rules. A variant that simplifies the first into a tidier version of itself adds no choice to the round. Send all three to the approval point together. Only a round arriving with no decision comp (a degraded roll, an identity-mode page, a direction pinned without the decision round) renders all three here.
-- A comp is a designed surface, not a picture of the subject. Lead the prompt with the surface's own structure: the regions this design has, named in order with their scale relationships. Write it in the order the visitor reads it: the headline in its own words and the primary action first, then the one focal move in full detail, then each remaining region in a phrase; the image model fills whatever the prompt describes in detail, so a region you inventory renders as a second focal move. A page with no navigation says so instead of inventing one, and an unconventional surface states its unconventional skeleton. A prompt that leads with atmosphere gets a vignette back: the model paints the fish market instead of the fish market's website. How far the world reaches into the page, and what it may carry, is the mode's call, in its COMPS rules. Render the page itself edge to edge: no browser window, address bar, device or desk around it, so prompt it as the page, never as a screenshot of a browser. Self-check every render: if it could hang as a poster, or reads as a photograph with some text on it, it is not a comp; regenerate with the dealt composition's regions named and its parts working in place.
-- The inverse is also a failure: a surface with none of its subject in it. The subject appears as the content the regions hold; the world dresses the frame and never displaces what the frame shows. The deletion usually rides in on the prompt's exclusion list, so exclusions bind invented claims, and a medium ban belongs to the committed imagery stance, never to caution. Before accepting a render, point at the subject; a render that depicts everything about the world and nothing of the subject fails however faithful its atmosphere. Regenerate with the subject's content named region by region.
+- A comp prompt is short, 150 to 300 words, and carries only what the image model cannot know. The model draws what the prompt names and stops deciding whatever the prompt pins down, so a prompt that specifies everything gets back a template. Open with the surface: the kind of screen it is (a desktop web app, a documentation page, a landing page, a native app screen at its device viewport) for the product, by name, rendered edge to edge with crisp readable text and no browser window, device, or desk around it. Then the product and the visitor's job in plain words, with the real content that answers it: names, numbers with units, the offer or the state, any claim at its honest certainty, and the next action by name. Then what leads this comp, its product concept (summary-led, timeline-led, offer-led, reference-led), and the parts the screen holds in reading order, a phrase each. Then the frame in one line: the product's mark and name, its navigation, its utility controls, and the far edge, so the comp is a finished screen, never a fragment; a page with no navigation says so instead of inventing one. Leave sizes, positions, colours, and component styling out, except a colour, typeface, or brand element the user pinned: that is content, so state it. Self-check every render: if it could hang as a poster, or reads as a photograph with some text on it, it is not a comp; regenerate.
+- The look comes from the world, and from a picture wherever one exists. An established world has the screenshot reference above: it carries the look, so the prompt adds one sentence on what the reference contributes and nothing more about style. A catalog world has card images: attach its board and hero as reference images (the NEXT line names them; `impeccable generate-image --ref <board> --ref <hero>`, or the harness image tool's input images) and write one sentence, that the references are style inspiration adapted into a real working product, their layout and content not to be copied. Describing the style as well overrides the picture. A direction with no card images carries its look in two to four sentences of concrete visual law: the ground and main colours with what each does, the type voices, and the one signature that makes it recognisable. No path lists colours past the main few or describes components one by one.
+- The subject must be in it. Before accepting a render, point at the subject: a render that depicts the world and nothing of the subject fails however faithful its atmosphere, and the deletion usually rides in on an exclusion list. Regenerate with the subject's content named.
- Judge a comp as the shipped screen: the visitor's job must be readable from the image alone. Name the surface's mode from the render with no caption; a render whose mode cannot be read back is art direction without a surface. Regenerate with the visitor's job as the prompt's spine.
-- Commitment is depth, not performance spread across regions. The world enters through one dominant move plus the material, type, and spacing that support it; the remaining regions hold still in that material so the move can be read. A region that simply does its job in the world's grammar carries the direction further than a region performing the concept. The check cuts competition, never content: a quieted region keeps its information and stops performing, and the page still carries its content (a way to navigate, the offer, its supporting copy, an action, a glimpse of what follows) at real reading size (body text near 16px on a 1440px-wide page). That is a list of content, not a layout: where each piece sits is the composition's call. A second element competing with the named focal moment at the same scale means the comp is shouting; with no named focal moment, several regions performing the concept at once is the same shout. Regenerate keeping the strongest move and quieting the rest. Busy is louder, not bolder.
+- Commitment is depth, not performance spread across regions. The world enters through one dominant move plus the material, type, and spacing that support it; the remaining regions hold still in that material and keep their content at real reading size. A second element competing with the lead at the same scale is the comp shouting: regenerate keeping the strongest move and quieting the rest. Busy is louder, not bolder.
- When the user shortlisted multiple concepts, spread the three across them.
-- When one direction is committed, vary the structural uncertainty an image can resolve: topology, sequence, density, hierarchy, focal composition, or interaction framing.
+- When one direction is committed, vary the structural uncertainty an image can resolve: what leads the screen, topology, sequence, density, hierarchy, focal composition, or interaction framing.
- Show enough beyond the opening moment to prove the concept can govern the whole surface.
- Do not generate a palette artifact, ask new atmosphere questions, introduce a different type voice, or invent a new motif. If the committed world cannot support the concept, return to the concept shortlist rather than changing the world.