Configure Pretext typography settings for PDF export — font stacks, sizing, line height, page geometry, and disposition-driven typography mapping. Use when defaults need tuning for specific content types or audiences.
Install with Codex or Claude Copy this prompt, paste it into Codex, Claude, or another assistant, and let it review the skill page and install it for you.
A direct command skips the review prompt. Inspect the source before running it.
Configure Pretext typography settings for PDF export — font stacks, sizing, line height, page geometry, and disposition-driven typography mapping. Use when defaults need tuning for specific content types or audiences.
model
sonnet
category
pdf
triggers
["configure pretext typography","tune PDF typography settings","adjust font stack for PDF export","set PDF page geometry"]
tier
1
agents
["primary"]
tool_dependencies
["file_system"]
inputs
[{"name":"content_type","type":"string","description":"Content type to tune typography for — prose, code, chat, or structured","required":true},{"name":"typography_overrides","type":"object","description":"Typography settings to override (font_stack, body_size, line_height, page_size, margins)","required":false}]
outputs
[{"name":"typography_config","type":"object","description":"Resolved Pretext typography configuration object ready for use in pdf-export"}]
PDF Typography Skill
Version: 0.1.0
Created: 2026-04-08
Author: Dojo Genesis
Purpose: Fine-tune the Pretext typography engine for optimal PDF output across different content types, audiences, and aesthetic goals.
I. The Philosophy: Measure Twice, Render Once
Pretext does one thing exceptionally well: it measures text on a real Canvas surface, then reports exact dimensions. This means every line break, every paragraph spacing, every margin is computed from actual glyph metrics — not CSS approximations. Typography configuration is the lever that controls how this precision manifests in the final PDF.
II. When to Use This Skill
Content-specific tuning: Code needs different settings than prose. Chat needs different settings than reports.
Audience adaptation: A client-facing PDF needs different typography than an internal code review.
Brand alignment: Match typography to organization design tokens or brand guidelines.
Disposition mapping: Map ADA agent personality traits to typographic choices (DojoChat export).
III. Steps
Step 1: Identify Content Type
Each content type has an optimal typography profile:
Content Type
Body Font
Code Font
Size
Line Height
Max Width
Prose (default)
Inter
Recursive
11pt
1.5
468pt (6.5in)
Code review
Inter
Recursive MONO=1
9pt
1.3
504pt (7in)
Chat export
Inter Variable
Recursive
13pt
1.45
420pt (5.8in)
Presentation
Inter
Recursive
14pt
1.6
432pt (6in)
Reference manual
Inter
Recursive
10pt
1.4
468pt (6.5in)
Step 2: Configure Font Stack
The Pretext bridge supports three font families with variable axes:
A TypographyConfig object ready for consumption by the Pretext prepare()/layout() pipeline, plus page geometry settings for the PDF renderer. Concretely: validated font descriptors, line height values per content zone (body/code/headings), page margins in points, and column count. When used with pdf-export, the output becomes embedded in the rendered PDF — font choices and spacing are baked into the byte stream, not post-processable.
Examples
Scenario 1 — Client-facing Go spec PDF (mixed prose + code):
Content type: prose primary, code secondary. Audience: non-technical client.
Body: Inter 12pt, opsz=16, line height 1.6
Code: Recursive MONO=1, 10pt, line height 1.35
Headings: Fraunces 700, SOFT=30, WONK=1
Page: Letter, 72pt margins, single column
Validate: body line ~58 chars (within 45-75 target)
Code: Recursive MONO=1, 9pt, line height 1.3 (same size as body — intentional)
Headings: Inter 700, no Fraunces
Page: Letter, 54pt margins, potentially 2-column
Validate: code blocks fit within 504pt max width without overflow
Edge Cases
Canvas vs. DOM font resolution on macOS:system-ui resolves to different faces in Canvas and DOM contexts. Always pass explicit family names (Inter, Recursive, Fraunces) to fontDescriptor.
Optical sizing mismatches at large scales:Inter at 48pt (heading) with opsz=16 (body default) produces thin, fragile letterforms. Set opsz to match intended display size.
Mixed content with extremely long code lines: If source code has 120+ char lines and the page is letter-width, either reduce font size or switch to landscape. The layout engine will not auto-truncate — overflow is clipped.
Recursive CASL axis in monospace mode: Setting MONO=1, CASL=1 simultaneously produces undefined intermediate rendering on some PDF engines. Use CASL=0 when MONO=1.
Two-column layout with tall code blocks: Page-break logic does not split code blocks across columns. A single tall block will overflow into the gutter.
Anti-Patterns
Specifying system-ui in any Pretext context. Breaks cross-machine reproducibility and causes Canvas/DOM measurement drift. Use explicit font family names.
Ignoring opsz at heading sizes. Using body opsz (14-16) for large headings (24-48pt) produces optically incorrect letterforms. Set opsz to match render size.
Setting line height below 1.3 for code. At 1.0-1.2, indentation guides and leading dots between columns become invisible. Minimum 1.3 for monospace content.
Double-converting pt/px units. Pretext measures in CSS pixels internally; the bridge converts to PDF points. Manually converting before passing to the API results in double-scaling (typically 72/96 = 0.75x error).
Applying a single TypographyConfig across all zones. Body, code, captions, and headings each need separate configuration. A one-size config either makes code unreadable or body text too large.
