AgentStack
SKILL verified MIT Self-run

Type Scale

skill-thrillmade-agent-skills-type-scale · by thrillmade

Use when designing, auditing, or extending a typography size scale built on a modular ratio. Names the canonical preset ratios (1.067 minor second → 1.618 golden), the picking heuristic by context (UI vs editorial vs marketing), the stops-up / stops-down convention (−2 to +8 default), the rounding rule (round to nearest integer px, rem derived after), and the upper-bound on ratios for product UI.…

No reviews yet
0 installs
1 views
0.0% view→install

Install

$ agentstack add skill-thrillmade-agent-skills-type-scale

✓ scanned · ✓ verified — works with Claude Code, Cursor, and more.

Security review

✓ Passed

No issues found. Passed automated security review. · v0.1.0 How review works →

  • Prompt-injection patterns
  • Secret / credential exfiltration
  • Dangerous shell & filesystem operations
  • Untrusted network calls
  • Known-malicious package signatures

What it can access

  • Network access No
  • Filesystem access No
  • Shell / process execution No
  • Environment & secrets No
  • Dynamic code execution No

From automated source analysis of v0.1.0. “Used” means the capability is present in the source — more access means more to trust, not that it’s unsafe.

Are you the author of Type Scale? Claim this listing to set pricing, connect Stripe payouts, and keep 70% of every sale.
Sign up to claim

About

Type scale

A modular type scale derives every font size from a single ratio applied to a base size. The discipline is choosing the ratio for the system's typical surface — UI density, editorial reading, marketing impact — then committing to it across the whole scale rather than per-role.

When to use

  • Designing a new typography scale from a base font size and a context (product UI / editorial / marketing).
  • Auditing an existing scale where sizes feel off-rhythm or step gaps look inconsistent.
  • Extending a scale up or down (e.g. adding a display rung above the existing max).
  • Code review: flag hand-picked sizes outside a modular ratio, ratios above 1.5 in product systems, or rems-as-source-of-truth (always derive rem from rounded px, not the other way around).

When NOT to use

  • Editorial-only books / long-form prose where the ratio is a typographer's pick and a modular scale would over-constrain. The discipline still applies to documenting the choices, not to forcing a ratio.
  • Single-size systems (rare; usually a stopgap during initial system definition).

Canonical preset ratios

| Ratio | Name | Best for | |---|---|---| | 1.067 | Minor second | Very dense UI; gaps so small that most stops feel near-identical | | 1.125 | Major second | Dense UI (admin consoles, data-heavy tools); subtle hierarchy | | 1.200 | Minor third | A common default for product UI — balanced, readable, distinct headings | | 1.250 | Major third | Slightly more dramatic UI; comfortable for mixed product + marketing | | 1.333 | Perfect fourth | Editorial leaning; large gap between body and H1 | | 1.414 | Augmented fourth | Editorial; sharp hierarchy | | 1.500 | Perfect fifth | Marketing-leaning; aggressive but still usable | | 1.618 | Golden | Hero / display only; ratios above ~1.5 break down for product UI because step gaps grow too fast |

Upper bound: ratios above 1.5 rarely work for product UI — by step +5 the size is huge, and H4/H5 stop feeling distinct from body. Reserve 1.5 / 1.618 for editorial or marketing systems.

Stops-up / stops-down

A common default range is step −2 to step +8. Step 0 is the base size.

size(n) = round(base × ratio^n)
rem(n)  = size(n) / 16

Typical role mapping for a 16 px base, 1.200 ratio:

| Step | px | rem | Typical role | |---|---|---|---| | −2 | 11 | 0.6875 | Micro, legal, fine print | | −1 | 13 | 0.8125 | Small UI, metadata | | 0 | 16 | 1.0000 | Body, base | | +1 | 19 | 1.1875 | Lead paragraph, H5 | | +2 | 23 | 1.4375 | H4 | | +3 | 28 | 1.7500 | H3 | | +4 | 33 | 2.0625 | H2 | | +5 | 40 | 2.5000 | H1 | | +6 | 48 | 3.0000 | Display | | +7 | 57 | 3.5625 | Hero | | +8 | 69 | 4.3125 | Marketing splash |

The range is generous so a single scale spans dense UI through marketing without per-surface forking.

Rounding rules

Round font sizes to integer px by default:

  1. Compute the raw size: base × ratio^n.
  2. Round to the nearest integer px. Sub-pixel sizes render inconsistently across browsers and OS hinting.
  3. If two adjacent steps would round to the same px, bump the larger one up by 1 px to preserve monotonic spacing.
  4. Derive rem from the rounded px, not from the raw value. rem = rounded_px / 16. The other direction (round the rem) compounds with the px round and produces drift.

Rounding can be disabled for math-pure systems where fractional sizes are tolerated downstream. Document the choice if you turn it off.

Cross-references

  • REQUIRED BACKGROUND: line-height-grid — every size in the scale needs a line-height; the lh-ui (× 1.20) / lh-prose (× 1.50) tracks pair with this scale.
  • For role assignment after picking sizes: the typography-styles spec (compound tokens combining size + lh + tracking + weight + paragraph margin).
  • For the base size choice: 16 px is the common base; smaller bases (14 px) are valid for dense product UI but require revisiting line-heights and tap targets.

Verification

After picking a ratio and emitting the scale:

  1. Monotonicity: every step is strictly larger than the previous; no two adjacent steps share a px value after rounding.
  2. Round-trip: rem(n) × 16 = px(n) exactly (no drift from re-rounding).
  3. Distinguishability at small sizes: −1 and −2 are visually distinct (a ratio + base combo where they round to the same px is broken; pick a tighter ratio or larger base).
  4. Upper-bound sanity: if the +5 step is unusable for headings in your system, the ratio is too aggressive; drop to a tighter preset.

Sources

  • Modular Scale by Tim Brown — interactive picker for the canonical ratios.
  • Type Scale — same idea with live preview.
  • One worked instantiation: the UDTS typography foundations (its L2 skills are incubating in thrillmade/tokenomics).

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

Install and usage instructions live in the source repository linked above.

Reviews

No reviews yet — be the first.

Versions

  • v0.1.0 Imported from the upstream source.