Talk to the Director →
designesy.

Competitive benchmark

Benchmarks

Three tools, three questions. They do not compete — they answer different questions about design quality.

Hallmark prevents slop at generation time. Slop-eval scores slop after the fact. Designesy verifies contract conformance. This page maps what each catches that the others do not.

The three tools

Attributedesignesyhallmarkslop-eval
QuestionDoes this conform to the contracted design system?Does this look AI-generated?How much slop does this contain?
CategoryContract conformance verificationAnti-slop generation gateAnti-slop evaluation skill
DeliveryURL API + MCP tools (agent-invocable)Agent skill (prompt-encoded)Agent skill (deterministic script)
Checks36 across 13 categories58 binary gates + 6 pre-emit axes108 tells + 2 positive axes
LicenseSee designesy.orgMITApache-2.0
Stars20k40 (parent repo)
Score on designesy.org99.1% A (31/36 PASS)Not run (no URL API)Not run (no URL API)

They do not compete

The positioning

A complete design quality pipeline uses all three: hallmark to generate, slop-eval to catch slop, designesy to verify the contract. They occupy different positions on the design-verification spectrum — not the same position.

01

Generation → hallmark (prevent slop at emit time, 58 gates)

02

Evaluation → slop-eval (score existing designs, 108 tells + 2 positive axes)

03

Verification → designesy (verify contract conformance, 36 checks + MCP delivery)

Shared checks

Where the three tools overlap. The intersection is narrow — most checks are unique to each tool.

designesyhallmarkslop-eval
v06, v22 — Contrast (WCAG AA + APCA)gates 40-41X5 (critical)
v03 — Focus-visible ringsgate 26X14
v05 — prefers-reduced-motiongate 27
v11 — No transition:allgate 10
v28 — Reading width 45-75chgate 25
v26 — Font family count ≤3gate 37
v02 — No horizontal scrollgate 34
v24 — Touch targets ≥44px (WCAG 2.5.8)gate 26 (partial)

What designesy catches that they do not

20 checks with no hallmark or slop-eval equivalent. These are the uncontested layers — contract conformance, Unicode security, Core Web Vitals, forced-colors, AI disclosure, and the Cadence typography suite.

What they catch that designesy does not

designesy gaps. These are the AI-slop aesthetic detection, structural-variety, signature, and cohesion checks. designesy does not ask “does this look AI-generated” — it verifies contract conformance.

The moat

designesy’s uncontested layers

The layers no competitor approaches. These are not features — they are the category-of-one positioning.

Self-score on designesy.org

ToolScoreWhat it caught
designesy99.1% A (31 PASS / 0 FAIL / 1 WARN / 4 SKIP)v26 WARN: font family count (4, threshold ≤3)
hallmarkNot run (skill, not URL API)Would flag: pure #000/#fff (near-pure), zero-chroma neutrals (graphite is chromatic), nav structure
slop-evalNot run (skill, not URL API)Would flag: cool blue-charcoal dark (warm graphite, may pass), mono house voice (minor), focus states (should pass)

hallmark and slop-eval are agent skills, not URL-based APIs. They cannot score a URL without an LLM agent loading the skill and evaluating. designesy is the only one with a programmatic URL-based scoring endpoint — a structural advantage for CI/CD integration, leaderboards, and automated pipelines.

Provenance

Sources

Live research via AnySearch MCP (2026-07-30). designesy live score verified via POST to https://www.designesy.org/api/score (99.1% A, 36 checks). hallmark taxonomy from Nutlope/hallmark/skills/hallmark/references/slop-test.md. slop-eval taxonomy from fabricioctelles/skills/skills/slop-eval/references/tells.md.

Competitive benchmark — designesy vs hallmark vs slop-eval. The three tools answer different questions: hallmark prevents slop, slop-eval scores slop, designesy verifies contract conformance. A complete pipeline uses all three. /methodology · /leaderboard · /contracts/design-system