Typecraft Guide

0) When to use this skill

Use this skill whenever you need to:

• review typography quality in prose, UI copy, docs, slides, or style guides,
• convert rough copy into typography-correct copy quickly,
• generate platform-specific recommendations for text rendering.

1) Scope

The source playbook is in typography-reference.md. Source references live in the root README under [Sources](../../README.md#sources-with-publicationupdate-dates). This skill applies the following workflow to avoid over-correcting:

1. Identify platform + medium (web, iOS, android, print, presentation).
2. Pick goal: evaluation, improvement, or creation.
3. Apply the strongest rules first: body size, line-height, measure, contrast, and accessibility.
4. Report issues by severity (critical, important, recommended).

Reference loading policy

• Read typography-reference.md for nearly every typography task unless the request is only for a tiny stylistic rewrite.
• Read ../../README.md sources section when you need to justify a contested/platform-specific recommendation or include source-backed rationale.
• Read neither if the task is purely mechanical (e.g., “fix only one typo in body text already marked for correction”).

2) Operating rules

Core defaults

• Start with body typography first; all other styles derive from it.
• Keep confidence tags: [STRONG], [MODERATE], [CONTESTED], [PLATFORM].
• Prefer short, testable edits. Avoid style churn unless it changes readability or accessibility.

Severity definitions

• critical: violates legal/accessibility requirements or makes content hard to read.
• important: materially reduces readability, hierarchy, or consistency.
• recommended: improves polish, consistency, or maintainability.

Default evaluation sequence

1. Body size / line-height / measure.
2. Punctuation marks and special characters.
3. Emphasis (bold/italic/all-caps/small-caps/links).
4. Color contrast and hierarchy.
5. Spacing, tables, lists, platform constraints.
6. Accessibility and platform-specific checks.

3) Required output

For each pass, emit:

• Verdict: pass/fail/needs work.
• Top 3 findings with severity and exact rule.
• Before/after snippets where a concrete edit is beneficial.
• Short rationale tied to the platform context.
• Final check: contrast, scaling, and spacing tolerance.

When ambiguity exists, ask one targeted follow-up before making irreversible editorial changes (for example: copy tone constraints, brand style, legal typography requirements).

4) Platform and policy triggers

• Always read: [./typography-reference.md](./typography-reference.md)
• Always reference: [../../README.md#sources-with-publicationupdate-dates](../../README.md#sources-with-publicationupdate-dates)
• For strict code tasks, keep changes minimal and preserve existing copy intent.
• If a request is purely factual and asks for a rule table, provide the concise checklist and no narrative.

5) End condition

Stop when:

• all critical findings are addressed or explicitly deferred with rationale,
• the resulting result preserves original meaning,
• and accessibility thresholds are met (WCAG 2.2 requirements in the reference).