Layouts.

The grid is the input; the wordmark's anatomy is the converter. The X's optical centre sits at 0.8417 of the logo's width — that is where the X physically is, not a design choice — so choosing a grid line determines the logo's size: flush-left, w = (line − margin) ÷ 0.8417. The line breaks for the wordmark's clear space and runs margin to margin otherwise; when a mirror line answers it, the echo sits at the reflected grid line. Nothing is placed by eye: pick a line and an anchor, and the layout resolves.

Each document type carries its own grid, and every measure in this chapter lives on it: margins frame the canvas, gutters separate the columns, lines sit in the gutter centres, and the logo's sizes are solved from lines or spanned across whole columns. The source file's by-eye values confirmed the system — A4's measured line positions land on columns 4, 5, 6, and 8 of this grid to within a percent.

Content hangs off the line, not across it. A title block begins on the line the wordmark's X sits on and is capped at a grid line as its measure — so its position and width are grid decisions, not guesses; an eyebrow carries wayfinding above it, a subtitle below, and the engagement meta sits on the footer row. Type is sized as a fraction of the canvas, so it scales with the artifact. A template is a skeleton plus these slots, and its content is flat, serializable data — the exact shape a generated brief fills.

The report runs the standard 12-column document grid. Covers are quiet — a small mark on the column-3 or column-4 line — and scale up through the statement (column 8) to the full-span poster. The document header pairs its line with the echo at column 8; the back cover mirrors the cover at the footer row. Chapter 6:220's move is the strongest: sized from column 6, the X sits on the page's own centre line.

The same 12 columns at presentation margins. Working slides keep the mark small and cornered so the canvas stays free for content; divider and title slides let the line rule the frame. The deck system's five content layouts (title, section break, key points, statement, waterfall) sit on this same grid — these skeletons are the deck's identity moments: the opener, the dividers, the close.

The social tile runs a 20 × 20 pixel grid — 54px cells, margin of one cell. Lines sit on cell boundaries (6, 7, 8; the echo at 13), spans count whole cells (6, 18). The tile reads at feed size, so there are only three moves: the statement, the mark on the line, and the horizontal lockup.

The story carries the square's 20-cell module onto the tall canvas — same cell size, 36 rows. Every archetype here is proposed: the format doesn't exist in the source file yet, but the grid implies it, and the tall frame is where the line motif works hardest.

Twelve columns at tight margins. The logo is always vertically centred — it nearly fills the height, so only ticks of the line survive above and below the X. The banner's strongest move needs no line at all: sized from column 6, the X lands on the canvas centre and the line is implied.

The email header exists in the source file but carries no recoverable geometry — one hairline at the foot is the only trace. Both archetypes here are proposed: the document header at email scale, and the footer rule read as the horizontal lockup.

The geometry was measured from the Figma VI Overview file (.agents/vi-layout-spec.md holds the raw values and every anomaly), then snapped to the grids on this page — the normalisations, and the fact that proposed archetypes, formats, and every filled template await design sign-off, are logged in DESIGN-DECISIONS.md. Text slots ship now; imagery and pattern layers are the next generation behind these skeletons. Pack generation consumes the templates: it produces the flat content, the template owns the grid and the type.