posterly

name	posterly
disable-model-invocation	true
description	Build an academic conference poster (ICML/NeurIPS/ICLR/CVPR/etc.) as a single HTML/CSS file and render it to print-ready PDF via headless Chromium. Use when user says "做海报", "poster", "ICML/NeurIPS/ICLR poster", or asks to design/edit a research poster.
allowed-tools	Bash(*), Read, Write, Edit, Grep, Glob, AskUserQuestion, WebFetch, WebSearch

posterly — HTML/CSS Academic Poster Workflow

A poster is one HTML file styled for an exact print canvas, rendered to PDF via Playwright + Chromium. Iterate by measuring, not eyeballing — the screen preview lies; only emulate_media("print") at the correct viewport tells the truth.

Mental model

   HTML (with @page { size: W H })
     │
     ▼  print-emulate Chromium at W×96 × H×96 px viewport
     │
     ▼  data-measure-role tags identify columns/hero/footer-strip
     │
     ├──→ tools/poster_check.py measure  (HARD GATE — spread < 5 px,
     │                                    gap-to-strip ∈ [30,50] px,
     │                                    intercard gap ∈ [12,50] px,
     │                                    poster bbox aligns to page
     │                                    within ±2 px)
     ├──→ tools/poster_check.py preflight  (LaTeX residue, math `<`, missing imgs)
     ├──→ tools/render_preview.py  (PDF + thumbnail)
     └──→ tools/poster_check.py verify-final  (PDF page count / dims / size)

The skill is venue- and lab-neutral by default. Compose a design direction from templates/DESIGN-AXES.md (Step 2.5), scaffold from the nearest template in templates/README.md, edit :root design tokens to match the locked direction, fill TODO placeholders with your paper's content.

Canvas constants

Constant	Value	Notes
--u (CSS unit)	print = 1mm, screen = 1.6px	Use calc(N * var(--u)) for ALL sizing.
Print viewport (px)	W_in × 96 × H_in × 96	Computed by poster_check/render_preview.
Body cols	2 / 3 / 4, or 1 hero + 1 column	Per template.
Strict alignment	spread < 5 px (aim < 3)	Hard, non-negotiable gate.

Workflow

Step 0 — Pull the venue's official poster guidelines

Conference specs change year-to-year and vary wildly between venues:

• ICML often goes 60×36 in landscape; ICLR has been 24×36 in portrait in recent years; NeurIPS historically allowed multiple sizes; CVPR has used A0 portrait. Don't assume.
• Font minimums (≥24pt body for some venues), bleed margins, allowed orientations, on-poster logos, anonymity rules, QR-code policies — all vary.

Procedure:

1. WebSearch for "<venue> <year> poster instructions" or "<venue> <year> poster size".
2. WebFetch the venue's official page; extract dimensions, orientation, font-size floor, logo policy, anonymity rules, file-format requirement, template link if any.
3. If paywalled or down, check OpenReview's call-for-papers or ask the user for the relevant section.
4. Echo the extracted spec back to the user in one short table BEFORE drafting. Confirm before proceeding — a wrong canvas size invalidates every alignment decision downstream.

Step 0.5 — Design discovery (one round of AskUserQuestion)

Don't pick colors, logos, a QR target, the text density, or the block count silently. Ask the user in one round. (The template and the overall look are deliberately NOT asked here as a text question — they are decided in Step 2.5, where composed candidate directions are shown as rendered thumbnails; this round gathers that step's inputs.) AskUserQuestion takes at most 4 questions per call, so if more than four of the topics below need input for this poster, send the four most decision-relevant first and ask the rest (usually QR and block count) in a brief second call:

• Style leanings: "Any look-and-feel must-haves or vetoes? E.g. 'keep it light', 'a dark editorial look is welcome', 'no mascots' — or 'no preference'." Do NOT ask the user to pick a template or a style from a text list here: the layout skeleton and the whole visual direction are composed in Step 2.5 and chosen there by eye from rendered thumbnails. This bullet only collects constraints for that composition.
• Palette: "Lab/venue colors? E.g. #XXX accent + #YYY highlight — or say 'you pick'." When the user gives colors, use them as the palette seed. When they don't, do not silently fall back to the one house style: derive a poster-specific palette from the materials at hand (§Palette derivation below). Either way the palette is then shown, not just named — it lands in the Step 2.5 thumbnail candidates, where the user can veto it cheaply. The shipped neutral (steel-blue accent + warm-gold register) is the last-resort fallback, not the default.
• Logos & venue mark: "Any logos to place? Affiliation / lab logo, and the conference / journal logo — give paths or URLs, or say 'none'." Don't assume a venue logo is wanted; cross-check the logo policy from Step 0 (some venues forbid them). When logo files are provided, inspect each one (aspect ratio, transparency, background — Step 2 item 5) and pick a size class + chip treatment per Gate E — Header logos below; don't just drop them in at the default size.
• QR code: "Want a QR code? If so, pointing at which link — paper / arXiv / code repo / project page — or none?" Generate it offline as a local image (see Customizing in README / qrencode); never leave a remote QR-service URL in the poster — it hangs measure's networkidle wait and link-rots in print/archive.
• Text density: "How much text should the poster carry? (a) Normal (default) — posterly's usual concise balance of prose and paper figures; (b) Light — fewer words, with the saved space reassigned to paper-sourced figures/diagrams across the poster." For Light: trim secondary prose and merge or drop low-value text cards only when the freed area becomes visual real estate — larger AR-appropriate figures, figure-dominant cards, or additional useful paper visuals. Keep multiple figures while each stays legible; do not concentrate the budget into one enlarged centerpiece or switch layouts for that reason. "More room" means larger, clearer visual regions — never blank columns / cards / gaps: the Step 4 measure gate and the Step 6 anti-whitespace / figure gates all still apply.
• Block count: "How many content cards should the poster split into? (a) Normal (default) — the usual number of cards per column; (b) Fewer — fewer, larger cards for a calmer, less subdivided poster." This is orthogonal to text density: it controls how the content is boxed, not how much there is. For Fewer: consolidate related material into fewer, larger cards (merge adjacent cards that share a theme, fold a thin card into its neighbor) — keep every load-bearing section, number, equation, and figure; merge and enlarge, never delete substance, and don't shrink type to fit. Still fill the canvas — fewer cards means each card and its figures grow into the freed space; no blank columns / cards / gaps, the same measure and anti-whitespace gates apply.

Persist the user's answers as you go — re-reading them later prevents "improvement" loops that revert deliberate decisions.

Palette derivation (when the user has no color preference)

A paper already carries brand signals — the default palette should be derived from them, not house-styled. Pick the seed color from whichever signal is strongest for this poster (judgment call, no fixed priority):

• Affiliation brand color — the official identity color of the dominant lab/university (your own knowledge or a quick web check: Tsinghua purple, MIT cardinal, ETH blue…). Strongest choice when one affiliation dominates the author list.
• A provided logo — extract its dominant saturated color (snippet below).
• Venue identity — if the conference has a recognizable brand color.
• The paper's own figures — dominant hue of the headline figure; the poster then echoes its figures.
• Field/topic conventions — weakest signal; use only when nothing above gives a usable color.

Whatever the source, the seed feeds one fixed recipe — the rebrand surface is the same eight tokens in every template (--accent, --accent-deep, --accent-light, --accent-soft, --accent-ink, --emph, --emph-soft, --emph-ink):

from collections import Counter
from PIL import Image

def rel_lum(rgb):
    c = [v / 255 for v in rgb]
    c = [v / 12.92 if v <= 0.04045 else ((v + 0.055) / 1.055) ** 2.4 for v in c]
    return 0.2126 * c[0] + 0.7152 * c[1] + 0.0722 * c[2]

def contrast(a, b):
    la, lb = sorted((rel_lum(a), rel_lum(b)), reverse=True)
    return (la + 0.05) / (lb + 0.05)

def mix(rgb, other, t):  # t=0 -> rgb, t=1 -> other
    return tuple(round(v + (o - v) * t) for v, o in zip(rgb, other))

# 1) Seed. From an IMAGE (logo / headline figure): dominant saturated
#    mid-tone, bucketed so JPEG noise doesn't split the vote. From a BRAND
#    GUIDELINE: just set `seed` to the official hex and skip this block.
im = Image.open("images/lab-logo.png").convert("RGBA")
im.thumbnail((128, 128))
px = [(r, g, b) for r, g, b, a in im.getdata() if a > 128]
cands = Counter((r // 32, g // 32, b // 32) for r, g, b in px
                if max(r, g, b) - min(r, g, b) > 40       # saturated enough
                and 60 < (r + g + b) / 3 < 200)           # mid-tone
seed = (tuple(v * 32 + 16 for v in cands.most_common(1)[0][0])
        if cands else None)  # None = this image has no usable seed --
                             # try the next signal source, neutral only last

# 2) Tokens. Darken the seed until white text clears WCAG AA on it (the
#    same 4.5:1 also covers accent-as-text on white -- symmetric pair).
accent = seed
while contrast(accent, (255, 255, 255)) < 4.5:
    accent = mix(accent, (0, 0, 0), 0.08)
fmt = lambda c: "#%02X%02X%02X" % c
print(f"--accent: {fmt(accent)};  --accent-deep: {fmt(mix(accent, (0, 0, 0), 0.30))};")
print(f"--accent-light: {fmt(mix(accent, (255, 255, 255), 0.90))};  "
      f"--accent-soft: {fmt(mix(accent, (255, 255, 255), 0.82))};")
print(f"white-on-accent contrast: {contrast(accent, (255, 255, 255)):.1f}:1")
# --accent-ink stays #FFFFFF -- the AA loop above just guaranteed it.
# 3) Emphasis register: pick --emph per the rule below, then derive
#    --emph-soft = mix(emph, white, 0.90) and check --emph-ink (the ink
#    used ON the emph fill; template default #14314A) still
#    clears 4.5:1 against the register you chose -- swap it if not.

Rules that hold regardless of seed source:

• Print-safe accent: muted-to-medium saturation, medium-dark value. The AA loop above enforces the dark end; if a brand color is neon-bright, mute it toward the template's tone rather than shipping fluorescent ink.
• Emphasis register (--emph) is a per-poster choice, not a fixture: it is the single "ours / best" cue (the .ours row, ★ callouts, .keyword-emph), and defaulting it to the same color on every poster is a recognizable fingerprint. Pick ONE register per poster from a shortlist that suits the accent — warm gold #C9A24A (classic against cool accents), deep cool slate #3D4A5C (safe on any accent), rust #A2521C, forest #2D5F3E, burgundy #8F2437 (see templates/THEMES.md for the calibrated pool) — and vary the choice across posters. Constraints: (a) hue-distinct from the accent (rule 4 allows exactly these two hue families); (b) if the accent is warm (red/orange/yellow), the register must be cool; (c) re-derive --emph-soft as the register's ~90% white tint and keep --emph-ink at 4.5:1 on the register fill. (These constraints govern the default accent+emph role topology — a deliberately different Axis 3 choice made in Step 2.5, e.g. same-center tonal or categorical roles, follows templates/DESIGN-AXES.md instead.)
• Backgrounds default to near-white (--bg-page/--bg-card untouched, or at most a faint seed-hued tint) — this recipe derives the accent tokens, not the ground. A non-white canvas (cream / light tint / brand hue / near-black) is a legitimate Axis 2 choice made in Step 2.5, with its own contrast obligations (templates/DESIGN-AXES.md clash rules 6 and 9) and, for dark grounds, the "dark_ground": true declaration in the --tokens JSON (Step 2.5 item 4).
• Echo the choice: state the seed source and final tokens to the user (they surface visually in the Step 2.5 thumbnails) and record them in the Step 2.5 DESIGN DIRECTION comment block — "accent #660874 from Tsinghua brand; register slate #3D4A5C" — so a later edit doesn't "correct" a deliberate derivation back to neutral.

Step 1 — Confirm content & figures

With the venue spec and design-discovery answers in hand, ask once:

• Source paper path (paper-overleaf/.../main.tex ideal). Read the abstract, intro, headline results. Don't draft from memory — pull actual numbers, dataset names, equations.
• Figures: match images/ filenames to paper figures.
• Corresponding-author marker: which author gets ✉? Any starred (★) co-authors?
• Items to preserve/exclude: which sections to drop, any "do not revert" notes.

Step 1.5 — Content audit (mandatory; external reviewer recommended)

When to run it: this audits a filled draft, so do it once you've scaffolded (Step 3) and put real content into poster.html — but before you sink renders into the Step 4 measure/balance loop. It sits here, numbered with the content steps, because fidelity is a content concern, not a layout one: catching a wrong number now costs nothing, catching it after the layout loop wastes every render in between. The same audit repeats on the final poster at Step 6.5.

The draft must be audited for paper-to-poster fidelity. Past sessions caught real bugs ONLY here — paper said "20× fewer" but the table gave 16×, "fewest trajectories" was an overclaim vs the actual baselines, theorem preconditions were silently dropped. Skip this and you will discover errors only when standing next to the printed poster.

How to run it (in order of preference):

1. External LLM reviewer with file access (best). If you have Codex MCP, GPT-5 with file access, another Claude session, or any reviewer that can Read paper source files, use that. Recommended defaults if you have Codex MCP: model="gpt-5.6-sol", model_reasoning_effort="xhigh", sandbox="danger-full-access" (read-only audit — the sandbox often fails to start in containers / nested namespaces, and the audit only reads files anyway). Send the evidence pack + reviewer prompt below.

2. Self-audit (fallback). Walk every numeric claim on the poster and find its file:line in the paper source. Build the claim → evidence table by hand. Slower, easier to miss things, but better than skipping.

Evidence pack the reviewer needs:

1. The current poster.html (full)
2. Paper source path(s) so the reviewer can Read the .tex and any results/ CSVs
3. For every numeric claim, the paper file:line where the number originates
4. For every theorem/claim, the paper statement verbatim with all preconditions

Reviewer prompt template (use this verbatim, fill bracketed parts):

Audit the academic-poster draft at [poster.html abs path] against the paper at [main.tex abs path] (and any results in [results dir]). For every number, claim, theorem, dataset name, and method-comparison on the poster, produce a claim → evidence table:

  | claim on poster | paper file:line | paper says (verbatim) | match? |

Mark "match?" as: OK / NUMERIC-MISMATCH / OVERCLAIM / MISSING-PRECONDITION / NOT-IN-PAPER / SCOPE-NARROWED.

Then list every NON-OK row as a problem to fix before printing. Be skeptical — "all <method> methods" claims, "best by Nx" claims, and theorem statements without their epsilon/regularity preconditions are the most common silent errors.

You may proceed to Step 2 only after every finding is either fixed or explicitly recorded as "user-acknowledged tradeoff". Do not silently defer.

Step 2 — Image preprocessing (optional but reduces re-renders)

For each paper figure you'll use:

1. Vector source (EPS / PDF figure)? Chromium <img> renders neither EPS nor PDF (converting to PDF does not help — also not embeddable), so a vector figure must be converted first. SVG is best — it stays crisp at poster scale. If a vector converter is already installed (inkscape, pdf2svg, dvisvgm), go straight to SVG. If none is installed, ask the user (one AskUserQuestion) whether to install one for a sharp vector figure, or rasterize to PNG instead — don't decide silently:

   • Willing to install → SVG (preferred): e.g. inkscape fig.eps --export-type=svg, or pdf2svg fig.pdf fig.svg.
   • Decline → high-res PNG: rasterize with Ghostscript at ≥ 2× rendered px — gs -dSAFER -dBATCH -dNOPAUSE -dEPSCrop -r600 -sDEVICE=png16m -o fig.png fig.eps (PIL works too; it shells out to gs: Image.open('fig.eps').load(scale=5)).

   Never embed the .eps / .pdf directly — it renders blank, caught only late as polish's FIG/BROKEN after a wasted render.

2. Autocrop whitespace with PIL.ImageChops so the figure fills its card.

3. Re-export at ≥ 2× the rendered px — the print-quality target (the asset gate's hard floor is a lower 1.5×, so a 2× source clears it comfortably). A 200u × 120u figure print-rendered at 96 ppi → ~756 × 454 px. Source PNGs must be ≥ 1500 × 900 to look crisp at print.

4. QR codes: request at ≥ 2× rendered px (e.g., 480×480 if displayed at ~240 px).

5. Logos: inspect each user-provided logo file before placing it, then pick a size class and chip treatment from the two tables in Gate E — Header logos below. Use the same python that runs the posterly tools; this snippet needs Pillow (pip install Pillow if missing):

   from PIL import Image
   src = Image.open("images/lab-logo.png")
   w, h = src.size
   has_alpha = src.mode in ("RGBA", "LA", "PA") or "transparency" in src.info
   im = src.convert("RGBA")
   im.thumbnail((512, 512))  # analysis-only downscale
   tw, th = im.size
   px = im.load()
   edge = ([px[x, 0] for x in range(tw)] + [px[x, th - 1] for x in range(tw)]
           + [px[0, y] for y in range(th)] + [px[tw - 1, y] for y in range(th)])
   white_edge = sum(a > 240 and min(r, g, b) > 245
                    for r, g, b, a in edge) / len(edge)
   lum = sorted(0.2126 * r + 0.7152 * g + 0.0722 * b
                for r, g, b, a in im.getdata() if a > 32)
   p10, p90 = (lum[len(lum) // 10], lum[(len(lum) * 9) // 10]) if lum else (0, 0)
   print(f"AR={w / h:.2f}  alpha={has_alpha}  white_edge={white_edge:.0%}  "
         f"mark lum p10/p90={p10:.0f}/{p90:.0f}")
   

   Reading the output: AR drives the size class (Gate E table 1). white_edge >= ~70% on an image without alpha means a bare white background (Gate E table 2's "stray white rectangle" case). The mark's luminance percentiles — not the mean — say whether the marks are dark (p90 < ~120) or light (p10 > ~200); a white-filled logo with a thin dark outline fools a mean. An SVG logo can't be opened by PIL — parse its viewBox for the AR and judge the chip from the rendered header crop in Step 5 instead.

Step 2.5 — Design direction (compose → thumbnails → lock)

Layout skeleton, canvas, palette, typography — every look-and-feel choice — is made here, as one composed direction, before any template is copied. The menu is templates/DESIGN-AXES.md (8 orthogonal axes, a devices pool, clash rules); the rendered option catalog is specimens/axes/index.html (one page per axis). This step sits after Steps 1–2 because two axes depend on knowing the content: density (Axis 5) is a capacity decision, and an Axis-1 focal choice needs to know the headline figure.

1. Concept first, then compose per axis. Start each direction from a concept statement — one line naming the world the poster lives in ("engineering blueprint — annotated schematic on grid paper", "midnight editorial", "archival index card"); the recipe names in DESIGN-AXES.md §Recipes are ready-made concept statements, free to adopt or adapt. Then for each of the 8 axes pick a primary option + modifiers (an axis choice is a structured object, never a bare enum pick), plus 0–2 devices from the pool — each pick derived from the concept: if you can't say in one phrase how a pick serves the concept, it's decoration — swap it for one you can, or default that axis to quiet. A merely-legal combination that serves no concept is exactly the "assembled, not designed" look this step exists to prevent. Feed in the Step 0.5 answers: user/derived colors → the Axis 3 seed (§Palette derivation); text density and block count → Axis 5; style vetoes → hard constraints. Then walk the clash rules at the bottom of DESIGN-AXES.md: check all 9 hard rules one by one against the composed set (check, don't debate); a soft rule you trip stays legal, but write the tradeoff down in one line — e.g. "cream canvas + grotesque type: accepted, the letterspaced eyebrows carry the editorial tone". Finally, every direction designates its hero moment — the single loudest element on the sheet (an oversized headline number, a statement masthead, a dominant hero figure, one full-bleed band; usually the Axis 1 focal choice or one device doing double duty). Exactly one: two competing loud elements read as noise, zero reads as an unfilled template. Everything else sits at least a register quieter, and in the Step 4–6 space fights the hero moment is not the first thing you shrink.

2. Compose 2–3 candidates, far apart. One direction is a proposal, not a choice — compose 2–3 so the user actually chooses. Candidates must be distinguishable at thumbnail size: every pair must differ on at least two of the four fingerprint axes — canvas base (Axis 2), frame-line (Axis 6), section-heading joint (Axis 7), masthead (Axis 8). Two candidates that differ only in accent hue are the same candidate twice. Tag every candidate with its build cost before showing it: (a) current templates/components realize it directly; (b) it needs a construction ported from the specimens/axes/ catalog (the token-native CSS exists but must be adapted to the poster's tokens and units, not pasted — normal); (c) it needs a brand-new system component — offer it only with that caveat, and a user pick of a (c) candidate is a direction preference, not yet the lock: raise the Step 6 escape-hatch system-extension proposal first, and lock/scaffold only once it's approved. Never show the user a thumbnail you can't build.

3. Thumbnail pre-selection. For each candidate build a quick style specimen — NOT a full poster: one small HTML file with a masthead bar plus one column of 2 cards, applying that direction's canvas, palette, typography, frame-line, and section-heading joint (placeholder copy is fine; no figures, no @page, no data-measure-role). If the direction's hero moment is a visible structure (statement masthead, giant stat, full-bleed band, dominant figure), sketch it as one abstract placeholder block at roughly its true relative scale — a flat grey box stands in for a figure — so the user isn't picking the loudest element blind. Render each to a small PNG with Playwright — same file-URI → screenshot pattern as tools/render_preview.py, only simpler: a fixed viewport (~800×1000 px) and page.screenshot(), no print emulation, no PDF. Specimens are throwaway working files (direction_a.html / direction_a.png in the work dir — never in templates/) and run no gates: no run_gates.py, no measure, no polish — they exist only to be looked at.

   • Interactive session: show the PNGs, then ONE AskUserQuestion — one option per candidate (short label + the candidate's concept statement and hero moment as its one-line sketch) plus "none of these — recompose". For an (a)/(b) candidate the pick is the lock; a (c) pick is a direction preference — run the Step 6 escape-hatch system-extension proposal first and lock only on approval.
   • Non-interactive session: lock the recommended candidate yourself and state why (fit to content volume, venue tone, wave-level anti-convergence) in your report.

4. Lock and record. Record the locked direction as an HTML comment block that goes at the top of poster.html the moment Step 3 scaffolds it (and stays there through every later edit) — axis by axis, with modifiers, devices, and the anchor poster IDs you leaned on:

   <!-- DESIGN DIRECTION (Step 2.5, locked)
     concept:        cream editorial -- a journal front page: serif masthead, framed plates, rust accents
     axis1 layout:   3-column, top-hero (focal: pipeline figure)     [62396, 66579]
     axis2 canvas:   cream, flat, full                               [64736]
     axis3 palette:  accent+emph; accent #0F6070 (lab-logo seed), emph rust #A2521C
     axis4 type:     serif display + sans body; small-caps eyebrows  [64736]
     axis5 density:  balanced
     axis6 card:     white surface, thin colored frame, shadow       [65714]
     axis7 heading:  underline rule + number chip                    [65287, 63757]
     axis8 masthead: brand band; footer: contact strip               [63030]
     devices:        metric scoreboard, QR CTA
     hero moment:    top-hero pipeline figure -- the one loud element; scoreboard stays a register quieter
     soft-clash notes: none tripped
   -->
   

   Then write the tokens pack — always, named design_tokens.json, next to poster.html, holding at minimum the accent/emph hue centers — and pass it on every gate run as run_gates.py … --tokens design_tokens.json (forwarded to style_check.py):

   • hue centers ("hue_centers": {"accent": <deg>, "emph": <deg>}) — rule 4 reads exactly these two slots and allows at most two non-neutral hue clusters. Dual-semantic: map the two semantic hues onto the accent/emph slots and a re-enabled rule 4 checks clean. Categorical (3+ hue roles): rule 4 cannot pass and must stay disabled (it is off by posterly default, --style-disable 4,5) — record the full palette in the DESIGN DIRECTION block instead;
   • vendored font families ("fonts": {"serif": […], "sans": […], "mono": […]}) when the Axis 4 voice is off the built-in whitelist (vendor the files locally, never a CDN);
   • "dark_ground": true iff the Axis 2 base is near-black or a dark brand hue: it switches off style rule 12's large-dark-area warning, which is calibrated for light posters and would otherwise fight a deliberate dark ground. Never set it on a light poster to silence a rule-12 warning about an oversized dark slab — there the warning is reporting a real problem.

Anti-convergence. The shipped default (white canvas · soft card · plain headings · centered masthead) is one combo among many, not the home position: landing there after a fresh composition is fine; landing there every time is a fingerprint. In a wave (several posters in one batch), consecutive posters must differ on at least two of the four fingerprint axes (canvas / frame-line / section-heading joint / masthead) and must not reuse the previous poster's concept statement — read the previous poster's DESIGN DIRECTION block before composing the next, and hold the --emph register decision at wave level per templates/THEMES.md Mechanism 1.

Step 3 — Scaffold from the gallery

1. cp templates/<chosen>.html <work-dir>/poster.html — <chosen> is the template whose skeleton is nearest the locked direction's Axis 1 topology (templates/README.md table); the remaining axes are applied on top as token edits and component swaps.
2. Paste the DESIGN DIRECTION comment block (Step 2.5) at the top, then edit the :root design tokens (single block; affects everything) to realize the locked direction — palette (Axis 3), typography (Axis 4), density scale (Axis 5), frame/radius tokens such as --rs (Axis 6). The figure mount belongs to that same Axis 6 decision: restyle --fig-bg / --fig-frame with the cards so paper figures sit in the design instead of pasted on it (transparent-PNG ground + keyline; captions already run on --text-secondary, block-figure caption <strong> additionally on --accent-deep).
3. Replace <title>, header (title/subtitle/authors/affiliation), banner (if any), column cards, takeaways strip (if any), footer.
4. Match the template's data-measure-role scheme — DO NOT remove these attributes. The measurement script depends on them.
5. No logo / QR provided: keep the venue as its text badge — don't fabricate a venue logo. With no affiliation logo, delete the empty .logo-slot rather than leave a hollow box; the text affiliation line carries attribution. With no QR, delete .qr-block. Never fetch or invent an asset the user didn't give, and never leave a remote QR-service URL in the poster (offline local image only). (The corner .ornament watermark is an off-by-default Axis 8 identity accessory, shipped commented-out — enable it only when the locked direction calls for it.)
6. The takeaways strip is optional — judge it deliberately; don't default to keeping it or to cutting it. The landscape scaffolds ship with a bottom takeaways strip, but it earns its place only as a genuine 60-second narrative exit (3–4 one-line slots; the classic Idea / Method / Result / Practical labels are one example set — reword per the microcopy rule below). Keep it when it lands a conclusion the final column cards don't already; delete the whole .takeaways-strip block when those cards already close the argument or it would just restate the body — a redundant strip is worse than none (portrait templates omit it by design). When the poster is over-full — content fighting to fit, font sizes creeping toward the venue's floor, cards cramming together — this strip is the first block to drop to win the body its room back, and you should reach for that readily: cutting a merely-adequate takeaways row makes a better poster than shrinking everything to keep it. But the call is about content, not pressure: don't delete a strip that genuinely closes the poster just because space is tight, and don't keep one that isn't carrying its weight. Same spirit as "Fill means substance" below: a block stays only if it does real work. Hero exception (applies to this strip and the framework banner alike): a block that is the locked hero moment loses automatic first-cut status, and removing it — for any reason, content-merit or space — is a change to the locked direction, not a layout fix: in an interactive session ask the user to confirm the removal and the replacement hero (their pick was the lock); in a non-interactive one re-designate the hero yourself in the DESIGN DIRECTION block and record why.
7. The framework banner is optional too — same deliberate judgment, applied to the top. A poster does not have to open with a FRAMEWORK / TL;DR strip. Keep the .framework-banner only when the paper genuinely compresses to one sentence plus 2–4 headline numbers worth reading from 2 m. If the contribution doesn't reduce to a single line, or the opening is better carried by a hero figure (landscape_hero) or by the first column itself, delete the whole .framework-banner block and let the body grid absorb the height (then rebalance through the Step 4 measure loop). A banner that merely paraphrases the title or pads generic stats is noise at the poster's most valuable position — worse than none. When content is overflowing, this banner is likewise among the first things to cut — it holds the most valuable real estate for often the least load-bearing content, so reclaiming it for the body is usually the right trade and you shouldn't hesitate (a banner that is the locked hero moment follows the hero exception stated under the takeaways strip above: no automatic first-cut status, and removal only via user confirmation when interactive, recorded self-re-designation otherwise). Still, judge on merit, not pressure alone: a true one-line TL;DR with live headline numbers can be worth keeping even on a tight sheet. A method figure in the banner usually needs no caption — the banner's text block beside it already explains the method, so a figcaption just says it twice, and a long one is exactly what stretches the figure slot and strands the image with a dead band beside it (polish flags this as BANNER/IMAGE-SLOT). Default to a captionless banner-figure (<figure class="banner-figure"><img …></figure> — see COMPONENTS.md), never a hand-rolled .fb-fig or a bare <img class="w-100">. If a figure genuinely needs panel labels, bake them into the image or keep them to one short <figcaption> line (the component bounds the caption to the image width); centre a block image with margin-inline:auto, not text-align:center.
8. Microcopy is placeholder, not canon. The scaffold's small fixed words — the banner eyebrow, the takeaways strip title and slot labels, the QR label, the footer labels, the section names, the ★ key-marks — are stubs and examples, not house style: reword them to the poster's voice and concept (an "engineering blueprint" poster might label its exit strip "Field notes" and its eyebrow "Spec"). The old fixed set — eyebrow "Framework", strip "Takeaways: Idea/Method/Result/Practical", QR "Paper & Code" — shipping verbatim on every poster is a textual fingerprint, which is why those are now TODO stubs. Section names come from the paper's content: generic slots (Motivation, Method, Main Result) are fine anywhere, but the scaffold's distinctive ones (Key Insight, Why It Works, Numbers at a Glance) are examples to replace, not defaults to keep. In a wave, consecutive posters must not reuse each other's microcopy set.

Emphasis discipline (copy-level de-fingerprinting). Bold in body copy is earned per phrase, not budgeted. The test: must a reader 2 m away catch this in a 3-second scan? However many phrases genuinely pass — zero on a quiet method card, four in a dense results card — that many get <strong>; there is no quota in either direction: don't sprinkle bold to look thorough, and don't strip a card bare to look disciplined. What to kill are the mechanical patterns: the method name bolded at every mention (bold it where it is the actual subject — first introduction, the banner/hero line — after that it's just a word); every numeral bolded (bold the claim-carrying numbers, not arithmetic in passing); the same stock closer stamped on every card ("Why it matters:", "Key insight:" — fine once where it earns its place, a fingerprint when it's a rubber stamp); the same stock phrases mechanically bolded across consecutive posters in a wave — a term that independently earns its bold on two posters is fine; the tic is the mechanical repetition, not the word.

A gallery template is a scaffold: it passes preflight (structure) as shipped, but with figures commented out and copy as TODO stubs it is expected to fail measure/polish (columns only fill the top, so the column-bottom spread and gap-to-footer are far out of band). Those two gates judge a filled poster — they go green only after Steps 4–6 below, once you've added real content and balanced the columns. Don't try to "fix" a fresh scaffold to pass measure; fill it first.

Tools live in tools/ and read @page from the HTML, so they're canvas-agnostic — the same commands work for ICLR portrait and ICML landscape.

Theorem & equation sanity (quick, right after scaffolding). Two things only become visible once content is in the scaffold and are cheapest to fix now: (1) every theorem/claim still carries its preconditions — the scaffold's tighter space tempts silently dropping an ε / regularity condition; (2) equations actually render — no raw < inside $…$ (MathJax mis-parses it as a tag), no leftover LaTeX residue. preflight catches the mechanical cases; eyeball the preconditions. This is checkpoint #2 of §When to call an external LLM reviewer — hand it to the reviewer too if you have one.

Step 4 — Render + measure loop (HARD GATE)

Default driver: run_gates.py. After every layout change, run the whole sequence in one shot — preflight → style → measure → polish in load-bearing order (plus the asset gate only when you pass --manifest; otherwise it's reported NOT_RUN and excluded from overall), into one GATE_REPORT.json (see §Enhanced gates & fix discipline):

# After every layout change (the default loop driver). The Step 2.5 pack
# (design_tokens.json, always written at lock time) rides on EVERY call —
# dropping it silently un-declares your fonts / hue centers / dark_ground:
python <skill>/tools/run_gates.py poster.html --tokens design_tokens.json --report GATE_REPORT.json

Before the first loop iteration — run pack once (advisory). A column whose figures at their Gate A floors still overflow the footer-gap window — or at their ceilings still can't reach it — cannot be fixed by figure sizing at all, and discovering that inside the loop costs many wasted rounds. python <skill>/tools/poster_check.py pack poster.html probes both endpoints in the browser and names the column: REPACK_RECOMMENDED (move a card out / trim text before looping) or FIGURE_ONLY_UNDERFILL (the residual needs content, not figure growth). It is advisory (exit 0; floors are polish's WARN thresholds, not physical minima; hero panels aren't modelled) — treat it as the "should I re-pack cards across columns first?" answer, then enter the loop.

The loop is budgeted (script-enforced circuit breaker). measure counts consecutive failed measurements in an on-disk file next to the poster (.<filename>.posterly_budget.json, e.g. .poster.html.posterly_budget.json — survives context compaction); the first PASS, 12 h idle, or --reset-budget clears it. At the cap (default 30, --measure-budget, 0 disables) measure exits 3 with a CIRCUIT BREAKER banner and refuses to render again: stop iterating, re-think the layout (re-pack via pack, or reselect template/canvas) or escalate to the user with the current best state rendered — do NOT --reset-budget just to keep grinding the same edits. run_gates.py surfaces exit 3 as a measure FAIL and skips the remaining gates.

Work from the failure report, not the file. On a spread/gap/intercard failure, measure now prints (a) the shared passing band — the one bottom-range every column must land in — with per-column grow/trim ~N px [safe +lo..+hi] deltas, and (b) an edit targets block listing every card per column with its source line (L<n>), height, and a text anchor, marking the bottom card that sets the column bottom. Iterate from that report: jump to the source line or Grep the anchor, read the surrounding block to confirm you have the right card, edit, re-run. Don't re-Read the whole poster.html every round (the anchors are math-stripped locators, not verbatim source), and never emit the full file through your output (scaffold via cp, then surgical Edits). Full re-reads stay legitimate where they earn their cost: first contact with an unfamiliar/custom template, a cross-column re-pack, a structural/nesting failure, an anchor that's missing or ambiguous, and the final claim audit.

This is what wires the style hard gate into every iteration — the standalone measure call below does not run style. posterly runs style with rules 4 (≤2 hue families) and 5 (no gradients) disabled by default: palette and gradient choices are yours, while the rest of the design-system discipline stays enforced. Override with --style-disable '' to enforce all 13, or e.g. --style-disable 4,5,6,7 to also drop the font rules.

The standalone measure call is the minimum fallback — a quick single-gate spot check; it skips style/asset:

# Minimum / spot-check only (no style, no asset):
python <skill>/tools/poster_check.py measure poster.html
# Same single browser launch, plus the advisory polish report:
python <skill>/tools/poster_check.py measure poster.html --with-polish

--with-polish runs the polish measurement on the same rendered page (one Chromium launch instead of two) and prints its report at default thresholds. It is advisory there — it never changes measure's exit code; the loop's final soft gate remains a standalone polish run (--strict if you want it enforced).

Targets (defaults; configurable via flags):

• spread < 5 px across the last-card-bottoms of all columns (+ any hero panel). Aim < 3 px.
• gap to footer-strip/footer ∈ [30, 50] px — card shadow visible but cards don't float.
• intercard gap ∈ [12, 50] px — whitespace between consecutive stacked cards inside a column (side-by-side cards count as one row). The ceiling catches justify-content: space-between faking bottom alignment on an under-filled column: spread reads ~0 and the footer gap lands in band while a void sits mid-column (observed in the wild: 98–135 px voids against a 22.7 px design row-gap). The floor catches cards packed so tight the drop shadow (0 2u 6u in shipped templates) is buried under the next card, fusing the stack into one slab. Tune via --max-intercard-gap / --min-intercard-gap (floor 0 to disable for shadowless themes).
• position align ≤ 2 px (authoritative) — the [data-measure-role="poster"] bounding box must sit at (0, 0) to (viewport_w, viewport_h) within --position-tol-px. This IS the full-canvas requirement: a poster whose bbox aligns to the page is necessarily full-bleed. Catches transform: translate*, mis-positioned position: absolute, stray body margin in print, and CSS source-order cascade bugs where a screen rule wins over a print override.
• canvas-fill ∈ [95 %, 101 %] (coarse early diagnostic) — [data-measure-role="poster"] width/height ratio against the print viewport. Fires before the position check when the ratio is FAR off, with a more diagnostic error message that points at the common @media print { :root { --u: 1mm } } omission (renders at ~42 %) or hardcoded width > @page (renders at >100 %). For borderline 95–99 % cases, position-align is the truth. Tune via --min-canvas-fill / --max-canvas-fill. Safe-area design belongs as internal padding on a full-bleed .poster, NOT as a smaller poster — a smaller poster fails position-align.

This gate is non-negotiable. If measure exits non-zero, fix the layout — do NOT continue to render. Common fixes:

• spread > 5: shrink the column with the lowest last-card by reducing a paragraph's margin-bottom by 1u, trimming one line, or shrinking a fixed-height figure by 5u.
• intercard gap > 50: an under-filled column is being stretched. Remove justify-content: space-between/space-around from the column, use a fixed gap, and absorb the slack with CONTENT (grow a figure, add paper-sourced text per Gate C) — never with whitespace.
• intercard gap < 12: an over-full column is being squeezed by shrinking the row-gap, which buries card shadows. Restore the design gap (6u ≈ 22.7 px) and take the height back out of content instead (trim a paragraph, shrink a figure by 5u, or move a card to a shorter column).
• gap > 50 everywhere: body-grid is too tall; grow a card with substance (per Gate C / Fill means substance) or reselect a smaller canvas — don't leave the whitespace.
• gap < 30 anywhere: banner/header outgrew its slot; check .framework-banner rendered height.
• position misaligned (the usual full-canvas failure): make .poster full-bleed (width: 100%; height: 100%; margin: 0; padding: 0 in @media print); remove any transform: translate* or position: absolute offsets; ensure html, body { margin: 0; padding: 0 } in the print media query; and check that the print @media block comes AFTER the screen .poster rule so source-order cascade resolves the print override winning.
• canvas-fill < 95 % (diagnostic fired first): poster forgot @media print { :root { --u: 1mm } } so it renders at screen scale. Add the override.
• canvas-fill > 101 % (diagnostic fired first): hardcoded width: 1600px (or similar non---u-based size) exceeds @page. Replace with calc(N * var(--u)).

Fine-tuning levers — continuous vs. quantized. The fixes above move height in ~one-line jumps; the last few px to reach spread < 5 need a continuous lever, and not every knob is one:

• Figure width is continuous only when the figure is the column's bottom-most element — a centered/stacked figure, or a float tall enough that text never extends below it. In a float-wrap where text flows below the figure, widening it toggles whole text lines (one session: 48 % → 2823 px, 51 % → 3351 px — a 528 px jump for +3 %) and in the text-dominated regime it does nothing at all. Don't use figure width for sub-line alignment there.
• For a sub-line residual, add padding-bottom to the column's last card — continuous and zero-reflow (text doesn't re-wrap), and measure reads the card's border-box bottom so it raises the column cleanly. Lever of last resort, only for a < ~1-line residual on a normal-flow, auto-height last card (a flex:1 / fixed-height card won't grow this way). A large padding-bottom is a Gate-C smell, not this — it will (and should) trip CARD/TRAILING; fill big gaps with real content instead.
• line-height set on a .card won't reach its text — .card p / .card li carry their own line-height (higher specificity), so it silently no-ops. Override the text elements directly if you must compress line spacing.

poster_check.py measure also has these safety nets (so a false PASS shouldn't happen):

• Missing [data-measure-role="poster"] = hard fail.
• Empty columns = hard fail (override: --allow-empty-column).
• Missing footer-strip AND footer = hard fail (override: --allow-no-footer-gap).
• MathJax intended (a <script src="…mathjax…"> tag or window.MathJax config is present) but no <mjx-container> rendered, while TeX delimiters ($…$ / $$…$$ / \(…\) / \[…\]) remain in body text = hard fail (CDN block, script error). A page that just describes TeX syntax in prose without ever loading MathJax is NOT failed.
• MathJax typeset timeout = hard fail (override: --mathjax-timeout-ms).
• @page size missing AND no --canvas override = exit 2.

Run preflight in parallel:

python <skill>/tools/poster_check.py preflight poster.html

Catches: LaTeX residue (\ref{, \cite{, \textbf{, lone \ ), bare < inside $…$ math (MathJax mis-parses as HTML tag), missing local images, missing data-measure-role="poster", unknown role values.

Step 5 — Render + visual inspection

python <skill>/tools/render_preview.py poster.html
pdftoppm -r 100 poster_preview.pdf poster_check -png -f 1 -l 1
# then Read the resulting PNG

For dense regions, crop with PIL and read the slice — full poster at r=100 is ~6000 px wide; useful regions (header, banner, takeaways, one column) at full res reveal text wrapping issues invisible in the thumbnail.

Beyond defect-hunting, hold the render against its own DESIGN DIRECTION block once: at thumbnail size the locked hero moment should be the first place the eye lands (a competing loud element is a quiet-it fix in Step 6, not a redesign), and the sheet should still read as its concept statement rather than as parts from different posters.

Step 6 — Polish

After alignment is solid, run the visual polish gate:

python <skill>/tools/poster_check.py polish poster.html

This is a soft gate (exits 0 by default; pass --strict to fail on warnings). It surfaces failure modes that the hard alignment gate cannot see — figure sizing relative to its aspect ratio, typography orphans, column whitespace pretending to be balance, <br>-in-flex collapse, and header-logo problems (broken / oversized / QR-height mismatch / title squeeze). See §Visual polish gates below for the rule for each WARN class and the correct fix. Fix every WARN unless you explicitly judge it acceptable for this poster.

Other polish:

• text-wrap — match the property to the text:
  • balance only on short, centered display text (titles, captions, one-line takeaways ≤ 2 lines); it evens the ragged edge.
  • Never balance on multi-sentence prose (banner TL;DR, long takeaways), and especially not with text-align: left. Near a 2↔3-line threshold, balance shortens and hyphenates the first line to "even" the block, producing a crammed-left / big-gap-right banner. For prose that should fill its box, use text-wrap: pretty (fills each line, only protects the last-line orphan) or plain natural wrap.

Step 6.5 — Final review (strongly recommended): once run_gates.py is all-green and polish warnings are zero-or-waived, send the rendered PDF (or its high-res PNG slices) AND the HTML to the same kind of reviewer used in Step 1.5 (external LLM if available, self-audit otherwise). Same evidence-pack rule. The reviewer prompt focuses on four things distinct from Step 1.5:

1. Visual rhetoric: does the poster's narrative carry? Are the headline numbers prominent? Is the framework banner readable from 2 m?
2. Residue: any \ref{, \cite{, leftover TODO, raw < in math, missing image, broken QR link.
3. Final claim audit: re-check numbers and overclaims AFTER content has been polished — polish often introduces new claims ("a key advantage of…") that were not in the original draft.
4. Design coherence (judgment questions, not a restyle mandate — include the poster's DESIGN DIRECTION comment block in the evidence pack): Does the rendered sheet deliver its own concept statement, or do some elements read as pasted in from a different poster? Is the locked hero moment actually where the eye lands first, and is it the only loud element? Do figures sit mounted in the design — each component's native mount (ground / keyline / caption where the component has them) consistent with the direction — rather than dropped on top of it? Two component contracts the reviewer must not "fix": the hero-panel img is frameless by design (its stage frames it), and a banner figure is usually captionless. And do the small words and the bolding speak this poster's voice — or the scaffold's ("Framework" / "Takeaways" verbatim) and a formula's (the same words bolded on every card, one stock closer stamped throughout)? On emphasis there is no quota in either direction — zero-bold and several-bold cards are both legitimate; the question is only whether each bold earns its place. A miss here is almost always a small fix — retune a token, quiet one competing element, re-mount one figure — never grounds for a redesign. The locked direction and the venue's legibility floors outrank the reviewer's taste: in particular, discard generic "make it bolder" advice (bigger display type, louder colors, added ornament) that isn't answering one of these questions.

Fix every finding before declaring the poster done.

Step 7 — Final verification

python <skill>/tools/poster_check.py verify-final poster_preview.pdf \
    --canvas 60x36in --max-size-mb 20
# or read the expected canvas from the companion HTML:
python <skill>/tools/poster_check.py verify-final poster_preview.pdf \
    --from-html poster.html

Checks: page count == 1, dimensions match canvas, file size ≤ limit. --canvas accepts inch dimensions (60x36in) or named sizes (A0 portrait, A1 landscape). By default rejects swapped W/H unless the PDF declares Page rot ∈ {90, 270} or you pass --allow-rotated. --from-html <path> reads @page { size: … } from the HTML so they can't drift apart.

Then report to the user:

• File path of PDF
• Final spread (px) and gap-to-footer range
• Any unresolved Codex feedback
• Page-fit confirmation

Visual polish gates (Step 6 — soft gate)

Alignment passes but the poster can still look amateur. The failure modes below (Gates A–E, plus the framework-banner image-slot gate BANNER/IMAGE-SLOT) recurred across sessions enough to be promoted to first-class checks. tools/poster_check.py polish surfaces each as a WARN; the rules below explain how to fix.

Gate A — Figure sizing by aspect ratio

A figure too small for its column is more wasteful than one too big. Pick width by the figure's intrinsic aspect ratio (AR = naturalWidth / naturalHeight), NOT by a fixed default:

AR range	Shape	Aim for	polish warns below / above
AR > 1.3	Wide (workflow diagram, comparison chart)	90–100% of card width when the figure owns its card; 70–85% only when it genuinely shares the card with meaningful text	< 65% ⇒ FIG/WIDE (the defect FLOOR, not the target). Avoid image-left/text-right in a narrow column — text gets squeezed to nothing (deliberate, balanced exception: the data-fig-layout="beside-text" opt-out below).
0.8 ≤ AR ≤ 1.3	Square-ish (block diagram, scatter)	55–75% of card width	< 55% ⇒ FIG/SQUARE.
AR < 0.8	Tall (multi-series bar, long pipeline)	45–60% (centered if text-sparse, wrapped if text-rich)	< 36% centered ⇒ FIG/TALL-SMALL (small + side voids); > 70% ⇒ FIG/TALL.

The 90–100% aim for figure-dominant cards is field-calibrated: ResearchStudio's paper2poster ran a live poster wave at a 70% floor and still shipped visibly under-filled "small stamp in a big card" figures, then raised its gate to 90%-on-one-axis. posterly keeps the WARN floor at 65% because its cards legitimately mix text and figure (their cards are figure-dominant sections) — treat the floor as "definitely a defect below this", and the 90–100% band as where a figure that is the card's payload should land.

Thresholds are tunable via --wide-min-ratio / --square-min-ratio / --tall-max-ratio / --tall-min-ratio. The defaults bracket the documented "aim for" range, so a figure inside it passes cleanly. --tall-min-ratio (default 0.36) is a hard floor, not the ideal — a centered tall figure rendering below it (≈ a 64%+ symmetric side void) is the recurring "figure too small" bug; the ideal is still 45–60%. The floor is measured as rendered width ÷ card width, which runs a few points below the CSS width:% (card padding), so calibrate against the rendered figure, not the style value. Because polish is soft, a borderline figure you've consciously accepted can keep its WARN; raise the floor and run --strict if you want it enforced.

A figure whose <img> fails to load (missing file, 404, or unreachable remote URL) reports zero natural size and warns as FIG/BROKEN — it will be blank in print. An SVG legitimately reports zero intrinsic size, so it's exempt from FIG/BROKEN — but the AR sizing gates still apply to it, computed from its rendered aspect ratio (so a too-small/too-wide SVG figure is not silently exempt). The probe covers both card and hero-panel <img> (the hero centerpiece is the worst image to silently lose). One known gap: an SVG served from an extensionless URL still slips the FIG/BROKEN exemption heuristic (gets wrongly flagged broken).

Hero figures aren't exempt from sizing — HERO/STAGE-LETTERBOX. The card-width AR gates above (FIG/WIDE etc.) don't apply to a hero panel, but a hero figure can still waste its space: a narrow-aspect picture dropped into a wide-but-short .hero-stage is height-constrained and strands itself with big symmetric side voids (a 2:1 panorama height-capped in a 5:1 stage fills ~35 % of the width). HERO/STAGE-LETTERBOX fires when the picture fills < 55 % of the stage width while the stage is much wider (relative to the image AR) than the image needs, with symmetric voids. The usual root cause is cramming a second large figure into the hero so the main figure loses vertical budget — move the secondary diagram into a card / the supporting column, or constrain the stage width toward the image's aspect ratio. A genuine full-bleed hero (image AR ≈ stage AR, picture fills the width) never trips it.

A wide, short footer-strip can reclaim its title-row height — the vrail modifier. When a bottom band is wide and short and a horizontal title row eats height the body could use, move the title into a narrow left rail so the body takes the full strip height. This is the content-agnostic vrail modifier (COMPONENTS.md): it suits any wide-short full-width band — a row of small example figures (a benchmark / task gallery) is the common case (the figures enlarge), but any other wide-short band you author works the same — a metrics row, say. The .num badge stays upright on top and the title's words stack one per horizontal line (every word reads normally — never rotated/sideways), centered, which also evens out uneven word lengths. An over-long word is broken with a soft hyphen ­ at a sensible syllable point you (the agent) judge — not hyphens: auto, which no-ops in the headless renderer and lets the word overflow. Mark the title data-vrail-title so the deliberately narrow stack is exempt from the WIDOW check. Only on a wide, short, full-width strip — never a narrow column card, where the rail eats scarce horizontal width instead of freeing it (a title too long to fit without several hyphenations is the signal to shorten it or stay horizontal).

Concrete bad case (prior session): co-consideration.png (AR ≈ 1.41) shipped at 41 % column width. The whitespace beside it conveyed nothing and the figure was unreadable from 2 m. Fix: 66 % width, no text-right.

Deliberate image-left/text-right (the opt-out). FIG/WIDE fires on a wide figure sized below 65 % because the usual cause is the bad case above — a figure shrunk into a gray margin. But a wide figure that shares its card with a meaningful text column (figure left, explanatory annotations right) is a legitimate layout when the text genuinely earns its space and isn't squeezed to a sliver. For that case, mark the <img> with data-fig-layout="beside-text":

<img src="images/dynamics.png" alt="Training dynamics"
     data-fig-layout="beside-text" class="w-100">

This skips the AR width gates (FIG/WIDE / FIG/SQUARE / FIG/TALL / FIG/TALL-SMALL) for that image only — FIG/BROKEN still applies (a blank image is a bug regardless of intent). The attribute records the design decision in the markup, so a later edit (human or agent) reads "this figure is intentionally beside text" and leaves the layout alone instead of widening it to silence the warning. Use it only after you've eyeballed the render and confirmed the text column isn't squeezed — it is an opt-out for a verified-good layout, not a way to mute a real warning. examples/powerflow_icml2026/poster.html uses it on its training-dynamics card.

Center vs. wrap a tall figure (AR < 0.8). A tall figure is too small just as easily as too wide — shrunk to ~35 % and centered it's illegible with a big symmetric side void on each margin (the recurring bug FIG/TALL-SMALL now catches below a 36 % rendered width). Which layout to use is driven by how much text shares the card:

• Text-rich card → wrap the figure: text flows beside and below it, so the image can be large. Float it with .fig-wrap / .ff-fig (shipped in every template) and mark the <img> data-fig-layout="beside-text". The figure and its surrounding text must live inside one .fig-wrap — the clearfix is what makes the card grow to contain the float; a bare floated <img> escapes the card box and measure then reads the wrong card bottom.

  <div class="card" data-measure-role="card">
    <div class="section-title"><span class="num">N</span><span class="st-text">…</span></div>
    <div class="fig-wrap">
      <figure class="ff-fig w-50">
        <img src="images/arch.png" data-fig-layout="beside-text">
        <figcaption class="caption">…</figcaption>
      </figure>
      <p>…body text wraps to the left of, and then below, the figure…</p>
      <ul>…</ul>
    </div>
  </div>
  

• Text-sparse card → center the figure (.figure) at 45–60 %. There isn't enough text to wrap, so a float just leaves an L-shaped void; a centered figure at a healthy width reads cleaner. (Worked example, same poster: the architecture figure was wrapped — lots of surrounding text — while the Multi-Head figure was centered — sparse text. Opposite calls, driven by text volume, not by the figure.)

That L-shaped void is now a measured WARN: FIG/BESIDE-TEXT-VOID fires when the wrapping text stops more than 30 % of the figure's height short of its bottom (text too short to fill the float). Fix it in this order: (1) lengthen the text with paper-sourced detail until it fills the figure height, or (2) shrink the figure to the text height. But don't manufacture filler the section doesn't need, and don't shrink a figure below 2 m legibility just to match a short text block: if the figure is the point and the prose would be padding, convert the card to figure-dominant — drop the float, size the figure by its aspect ratio (Gate A, i.e. large), give it a one-line caption, and let the presenter speak the detail (see "Fill means substance", Gate C). That is different from the trap below: a Gate-A-sized figure that owns the card vs. a small figure centered inside a still-text-shaped card. Centering a beside-text float in place is the last resort, and only works for a WIDE figure (AR > ~1.3) — centering a square or tall figure at a width that fits a column just swaps the L-void for symmetric side voids and shrinks the figure, which is a worse look, not a fix. (Worked example: a square 2×2 surface-plot figure (AR ≈ 1.0) with a one-line caption was first "fixed" by centering — it landed tiny with big blank margins on both sides; the real fix was to keep it beside-text and add two paper-sourced sentences so the prose filled the figure's height.)

data-fig-layout="beside-text" is the opt-out marker for either layout (flex side-by-side or float-wrap) — it records "this figure intentionally shares its card with text" and skips the AR width gates. It is not a generic mute: it does not skip FIG/BESIDE-TEXT-VOID (a beside-text float still has to be genuinely text-rich — tagging a short-text float won't silence the void), and don't tag a centered, text-less small figure with it. Either enlarge it, wrap it, or (since polish is soft) accept the FIG/TALL-SMALL WARN.

Gate B — Typography orphans

measure checks card bottoms; it does NOT see a ↑ or superscript that wrapped alone onto its own line — a small but jarring artifact 2 m away.

Known orphan-prone patterns:

• Stat numbers with trailing arrows: 1.18–1.30× ↑, 18.24% ↓
• Number + unit / multiplier: 4096 × 4096, 7 nm, 2248×
• Trailing footnote markers: BraVE*, MEAN§

Defenses (preferred → fallback):

1. white-space: nowrap on the smallest element whose innerText must stay one line (.stat-num, .hs-stat .num, .takeaway-num). Blanket — text simply does not break.
2. Non-breaking glue between the last two tokens: 1.18–1.30× ↑. Use when nowrap would overflow a tight card.
3. Reword to put the marker first: ↑ 1.18–1.30× speedup — arrow becomes the line's first token, can't orphan.

polish flags an element ([class*="stat"], [class*="num"], .takeaway-num, .hs-stat, .headline-num) — emitted as ORPHAN — only when its text ends with a whitespace-separated trailing glyph from ↑↓↔×÷±§¶†‡*°% and lacks white-space: nowrap — i.e. a lone glyph token that could wrap by itself, like 1.18–1.30× ↑ or 18.24% ↓. It does not catch fused forms such as 4096 × 4096, 7 nm, 2248×, BraVE*, or MEAN§ (the text doesn't end on a space-separated glyph) — guard those by hand with the defenses above.

Lone wrapped fragment — display text that wraps to a short, stranded last line. The orphan logic extends from stat numbers to any short display text that wraps — a title, .section-title, banner headline, or one-line takeaway whose break strands a short fragment alone on the final line. The worst case is when that fragment is a word carrying a leading footnote/superscript marker (a *- or †-prefixed term): stranded and left-aligned, the marker+word reads as a stray symbol or a typo, not a word. polish does catch this now (WIDOW) — for the prose classes listed here. A centered heading/title is evened by text-wrap: balance, not this gate, so hand-check those — except the framework-banner body (.fb-text), which is centered yet is gated (it uses text-wrap: pretty, not balance) and held to the higher fill bar described below. It measures wrap geometry on .callout, .body-text, .caption, .section-title, .card p, .card li, .fb-text (framework-banner text) — including each <br>-delimited segment — and warns when the last visual line fills less than ~35% of the typeset width (the widest line of the block). Exception — the framework banner (.fb-text): as the poster's single most prominent text block it carries a much higher bar (~80%) — its last line must nearly fill the measure so the banner reads as a clean filled rectangle, not a ragged box. The fix is to reflow it to a near-full last line — use any one, or a moderate combination, of these parallel levers (no fixed order): (W) width — tune the .fb-text flex ratio against .banner-stats (the fill jumps non-monotonically with width — e.g. one real poster went 17%→97% between two nearby ratios — so try a few), without starving the stat boxes; (F) font size — bump .fb-text one --fs-* step, which (when it shifts the wrap) reflows the text and makes the most-prominent block bolder, without overflowing the banner or colliding with the stats; (E) expand — add a few truthful, on-message words so the last line grows to fill (no fabricated claims, and keep .fb-text under the ~400-char display cap or the gate stops measuring it); (T) trim — cut a few words so the block settles into one-fewer line that's all full and the runt disappears, without dropping a load-bearing number/term. Keep the change proportionate to the gap — don't push one lever to an extreme (a blown-up font, a starved stat strip) just to clear the %. Never force it with text-align: justify / text-align-last or letter-spacing padding (those strand ugly gaps / fake a rectangle the eye still rejects). After any fix, re-render and look — clearing ~80% is necessary, not sufficient. It judges by the last line's width, not its word count: a short two-word tail (= OMAD-only.) flags just like a one-word one, while a single long word that fills the line does not (its width ≈ the measure — it isn't stranded). Limits (still guard by hand): math/figures no longer hide a whole block — they join the line model as opaque cells. A last line trailing a figure/icon/table (img/svg/canvas/table), or one that is purely a lone equation (a trailing equation/figure with no real word, even with a sentence period — …by $\lambda$. has the word "by" and IS judged), stays unjudgeable; but a short text tail ending in inline math (mjx-container, e.g. …traded off by $\lambda$.) is caught — judged by its full visual width, math symbol included. It still skips white-space: nowrap/pre, RTL, elements marked data-vrail-title (a deliberately narrow vrail rail title — its short stacked lines and agent-chosen soft-hyphen breaks are intentional, not runts), and running prose over ~220 chars (display text .caption/.callout/.fb-text: ~400). Time widow fixes after measure is green. Rewording changes the paragraph's word count, which can change its line count — moving that column's bottom by a whole line-height and undoing alignment you already tuned (the same edit made before alignment costs nothing). Once the layout passes measure, prefer a rewording that keeps the block's line count (swap a word for a longer synonym rather than adding one) and re-run the gates after; balance and   glue re-break the existing lines without adding one, so they don't disturb alignment. Fix a non-banner WIDOW (or a fused-glyph case the gate can't see) by the option that makes the lines fill naturally, not merely the one that silences the gate — in this order of preference (the framework banner uses the parallel menu above, not this list): (1) reword — expand or contract the sentence — first choice for left-aligned prose (.callout, .body-text, .caption): move the break to a natural phrase boundary so the last line carries more of the measure and the line above fills to the margin. Adding one word often does it (e.g. "expressive" → "highly expressive"). (2) text-wrap: balance on a centered heading / title evens the lines so no fragment is stranded (the default on centered display text — keep it; never pair balance with text-align: left, per the wrap rules above). (3)   glue is a last resort: gluing the last two tokens (…local optima.) pulls the prior word down onto the last line and widens it above the threshold, so it now clears the gate only when it genuinely fills the line — but it still measures the last line's fill, never how full the line above is, so on left-aligned prose glue can still leave a ragged short line above with a big right-side gap. Glue is genuinely right where a token must not open a line: a leading footnote/superscript marker (… *term) or a tight stat cell. After any fix, re-render and look — a cleared gate is necessary, not sufficient. Never ship a stranded short last line, a lone * / † / ‡ fragment, or a line left half-empty by a glue "fix".

Gate C — Content-driven balance, not space-between-driven

justify-content: space-between on a column is a shortcut to bottom-align last cards across columns. It works ONLY when the cards' natural heights are within ~5% of each other. When they aren't, space-between fills the delta with empty pixels — usually one giant gap in the column with the smallest content.

Symptom: a column with one short card followed by 25 + mm of whitespace. Reads as "this column ran out of things to say".

Wrong fix: shrink gap globally to hide the whitespace. Peers still have meaningful internal gaps; reducing them makes the others claustrophobic.

Right fix: FILL THE SLACK WITH SUBSTANCE until the short column is within ~5 % of its peers' natural height — a larger paper figure where one earns the space, otherwise real paper content (see "Fill means substance" below for the figure-vs-prose order). Content you can recover from the paper:

• Sub-claims that were footnotes or implicit assumptions
• A 2–3 bullet "challenges" or "design choices" recap
• A short caption beside a previously-bare figure

Concrete bad case (prior session): the SnipSnap Motivation column shipped with a one-line "three challenges" summary, leaving a 13 mm space-between gap. Fix: expanded into 3 bullets matching the paper's challenge framing — column balanced via content, not whitespace.

Enforcement is two-layered. measure hard-fails any column whose gap between consecutive stacked cards exceeds --max-intercard-gap (default 50 px, absolute) — this is the backstop that catches the space-between shortcut regardless of mechanism (added after a production poster shipped 98–135 px voids with every gate green: spread read 0.00 px because space-between pinned the last card to the bottom, and the relative polish warn below stayed silent at 4–6 % of a 36-inch column). polish additionally warns earlier (emitted as SPACE-BETWEEN), when a column with computed justify-content: space-between has an inter-card gap exceeding 5 % of the column's height. Tune via --max-space-between-fill.

The same trap, one card. A single card set to flex: 1 (the standard way to make its column reach the footer and satisfy measure's spread/gap gates) is measured only by its bottom edge — a card stretched to twice its content's height passes measure with spread = 0 while the lower half is blank white. measure can't catch it (it checks only the bottom edge), so polish does: the CARD/TRAILING warning fires when a card leaves more than --max-card-trailing (default 10 %) of its height blank below its last line of content. A green bottom-edge gate is necessary, not sufficient. Never stretch a block to create whitespace just to make the layout "fit." Fix it like Gate C — fill the slack with substance (aim ≥ ~80 % full, not 46 %): a bigger paper figure first, real paper content when the figure can't carry it, per the figure-vs-prose order in "Fill means substance" below; if the section is genuinely that sparse, choose a smaller canvas instead — a single paragraph does not belong on a 60-inch sheet. A half-empty card reads as "ran out of things to say" and is a failed poster even when every gate is green.

The same trap, mid-card — CARD/INNER-VOID. A sibling failure mode: a row of equal-height cards (grid/flex + align-items: stretch) whose contents differ in height, where the short card pins its tail — a "Why it matters" footer, a takeaway line — to the bottom with margin-top: auto (or justify-content: space-* on the card). The taller card sets the row height; the short card stretches to match, and the slack opens as a band in the middle of the card — below the last real block, above the pinned tail. Because the tail still sits on the card's bottom edge, CARD/TRAILING reads ~0 and stays silent, and measure (bottom-edge only) passes. polish catches it as CARD/INNER-VOID: for every .card — not only the data-measure-role-tagged ones, so an agent-authored feature band is covered too — it measures the largest vertical gap between two consecutive stacked children and warns when that gap exceeds the card's stated row-gap by more than --max-card-inner-void (default 8 % of card height) and an absolute --min-card-inner-void-px floor (default 24 px, so a sub-line gap on a small card stays quiet). Side-by-side children (a flex row, a float) overlap vertically and never count — only a real vertical void registers. Fix it like Gate C: fill the short card with substance (a bigger figure first, then real paper content), or — when the cards genuinely differ in length — drop the bottom-pin / equal-height stretch so each card hugs its own content (the footer rows then no longer align across the row, but there is no void). This was a live miss: a 3-card "Main Technical Contributions" band whose middle card carried one equation vs two in its neighbours pinned its footer with margin-top: auto, opening a ~14 %-of-card void between the formula and the footer — every gate green, because the band's cards carried no data-measure-role (so CARD/TRAILING/measure never sampled them) and the void sat mid-card (so the bottom-edge checks saw nothing). The lesson for authoring: a content block — including one in a custom feature band — should carry the .card class so the void gates see it, and if a row of cards will hold unequal content, do not reach for margin-top: auto + align-items: stretch to fake a level footer row.

"Fill" means substance, not word-count — and a figure is substance. These anti-whitespace gates (Gate C, CARD/TRAILING, CARD/INNER-VOID, FIG/BESIDE-TEXT-VOID) exist to kill empty pixels, not to mandate dense prose. A poster is a talk aid, not a self-contained paper: you stand beside it, and the small details — a derivation step, a hyperparameter, an edge-case caveat — are yours to say out loud, not to cram onto the sheet. So the legitimate ways to fill a region are, in order: (1) a larger, more legible figure that earns the space; (2) figure-only, or figure + a one-line caption, when the figure already makes the point clearly — a self-explanatory plot does not need a paragraph restating it; (3) genuinely load-bearing prose. Reach for more text only when the figure can't carry the point alone. And figure legibility wins ties: if cramming text beside a figure would shrink it below what reads at 2 m, drop the text and let the figure be big (center it, short caption, presenter fills the rest) rather than starve the image to justify a paragraph. And if a region is genuinely that sparse — the figure is already as large as it should be and there is no real paper content left — the fix is a smaller canvas or dropping an optional block (banner / takeaways), never stretching whitespace to fill a sheet that's too big. What these gates fail is a half-empty card or a thumbnail figure marooned in whitespace — not a clean card whose work is done by one big figure and a few words.

Gate D — <br> line breaks inside a flex container

A <br> that is a direct child of a display: flex / inline-flex element is blockified into a flex item and stops creating a line break (CSS Flexbox spec — every in-flow child becomes a flex item). Intended multi-line content silently collapses: in flex-direction: row the "lines" lay out side-by-side on one row (often with a MathJax <mjx-container> baseline pulling one fragment up, so it reads as jagged "misaligned" text); even in column the <br> is a dead empty item. measure can't see it — the card bottom is unchanged — so it survives to print. Concrete bad case (prior session): an OPT-AIL banner loop label ↻<br>repeat<br>$K$ iters rendered as repeat and K iters jammed onto one row instead of three stacked lines.

polish warns LAYOUT/FLEX-BR when any flex/inline-flex element has a direct <br> child, reporting the computed flex-direction so the fix is obvious. Fix: wrap each line in its own <span> (or <div>) and set flex-direction: column with align-items: center / text-align: center; or, if the element doesn't need to be flex, make it a plain block where <br> works normally. Never rely on <br> for layout inside a flex box.

Gate E — Header logos (affiliation / venue) & title squeeze

Logos live in the header, outside any card or hero panel, so Gates A–D never see them. The failure modes are real and silent: a 404'd logo prints blank; a wide wordmark rendered at seal height becomes enormous and squeezes the title (the header grid is 1fr minmax(50%, auto) 1fr — the title sits in an equal-tracks-centred column floored at 50%; an oversized right block is caught by the right-block ratio, and either-side imbalance by the title-offset gate, below); a transparent dark mark vanishes on a dark header; a white-background JPG leaves a stray white rectangle on a colored one.

Sizing. The shipped .logo-slot uses a fixed height (so a low-res logo still upscales to target) plus a width cap with object-fit: contain as the extreme-AR safety net, and three size classes. Pick the class from the file's aspect ratio (the Step 2 logo inspection):

Logo AR (w/h)	Shape	Class on .logo-slot
< 0.7	Tall stacked mark	logo-tall
0.7 – 1.4	Square seal / crest	logo-square
1.4 – 2.5	Mid wordmark	(none — default)
≥ 2.5	Wide wordmark (university name banner)	logo-wide