Sample Gates in the Hallmark Slop Test: A Complete Guide to AI‑Generated Page Validation

The Hallmark Slop Test uses 58 binary yes/no gates to detect AI‑generated tells (slop) in web pages, with each gate targeting specific visual, structural, typographic, copy, or interaction patterns.

The Hallmark skill, developed in Nutlope/hallmark, validates every generated page against a comprehensive slop test defined in skills/hallmark/references/slop-test.md. Each gate is a strict binary check—yes means the page contains slop and must be revised before deployment. Below are representative sample gates that demonstrate how the test protects design quality across five architectural dimensions.

Visual Design Gates

Typography Palette Enforcement

Gate 1 in slop-test.md asks: "Is the display font Inter, Roboto, Open Sans, Poppins, Lato, or a system default?"

Any font outside this curated set flags a non‑catalog choice that may break theme consistency. This gate prevents arbitrary font selection that signals handcrafted deviation from Hallmark's controlled design system.

Gradient Restrictions

Gate 2 blocks specific color patterns: "Is there a purple‑to‑blue (or cyan‑to‑magenta) gradient anywhere—including a background‑clip: text gradient headline?"

Even atmospheric genres only permit background gradients. Text gradients are universally prohibited as decorative AI tells.

Structural Integrity Gates

Macrostructure Uniqueness

Gate 8 prevents template reuse: "Does the page reuse a structure it shouldn't—either the generic AI template or the same macrostructure as a previous Hallmark output in this project?"

This structural fingerprint check ensures each page differs from prior outputs unless the brief explicitly requests similarity. The enforcement lives at slop-test.md line 40.

Typography System Gates

Font Family Limits

Gate 37 enforces the Typography 2 + 1 rule: "Does the page use more than three distinct font-family families?"

The allowed configuration is:

• One display font
• One body font
• One optional outlier

A fourth family automatically triggers slop classification at slop-test.md line 98.

Content Quality Gates

Placeholder Detection

Gate 19 catches lazy copy: "Is there a placeholder name 'Jane Doe / John Smith' or a startup cliché (Acme, Nexus, Seamless, Unleash)?"

Generic placeholders make pages read as templates rather than bespoke designs. This gate appears at slop-test.md line 54.

Metric Honesty

Gate 46 enforces factual accuracy: "Invented metric—any quantitative claim the user didn't supply (e.g., '10× faster', '50 000+ teams')?"

Fabricated numbers are a critical slop gate. The required fix is replacing the number with a placeholder or requesting real data from the user.

Decorative Element Gates

Semantic Anchor Requirement

Gate 45 targets meaningless ornamentation: "Does the hero contain a decorative element (cursor, scanline, gradient blob, abstract shape, ornament, badge, sticker) that has no semantic anchor in the content?"

Decorative assets must tie to content—a cursor indicates a command, a scanline implies terminal context. Pure decoration without meaning is slop at slop-test.md line 138.

Layout Pattern Gates

Eyebrow Placement Rule

Gate 54 eliminates a common AI tell: "Section eyebrow / tag beside the heading (tag‑left, header‑right)."

The classic left‑margin label pattern reads as templated editorial output. Hallmark requires the eyebrow stacked above the heading, never side‑by‑side—enforced at slop-test.md line 180.

How the Slop Test Executes

During a Hallmark run, the skill implements a pre‑emit self‑critique across six axes. If any axis scores above threshold, the page fails validation. The 58 gates cover:

• Visual — colors, gradients, imagery
• Structural — page architecture, template reuse
• Typographic — font selection, pairing rules
• Copy — placeholder text, invented metrics
• Interaction — decorative elements, motion patterns
• Layout — grid patterns, component placement

Each gate references specific line numbers in skills/hallmark/references/slop-test.md for traceable enforcement.

Summary

• The Hallmark Slop Test contains 58 binary gates that detect AI‑generated tells
• Gates are organized across five dimensions: visual, structural, typographic, copy, and interaction
• Typography gates enforce the 2+1 font family rule and curated display fonts
• Structural gates prevent template reuse and macrostructure duplication
• Content gates block placeholder names and invented metrics
• Decorative gates require semantic anchors for all visual elements
• All gates are defined and traceable in skills/hallmark/references/slop-test.md

Frequently Asked Questions

What happens when a page fails a slop test gate?

The page must be fixed before shipping. Each gate is binary—any yes response indicates slop presence. The Hallmark skill returns specific gate references from slop-test.md so developers know exactly which pattern to correct.

Can slop test gates be customized or disabled?

The source analysis does not indicate configurable gate sets. The 58 gates in skills/hallmark/references/slop-test.md appear to be standardized across all Hallmark runs, ensuring consistent quality enforcement regardless of project.

Why does Hallmark use binary yes/no gates instead of scoring?

Binary classification eliminates ambiguity. A scoring system might allow "acceptable" levels of slop, which contradicts Hallmark's goal of shipping pages indistinguishable from human-crafted designs. Yes equals slop, period.

Where can I review the complete slop test specification?

The full 58‑gate specification lives at skills/hallmark/references/slop-test.md in the Nutlope/hallmark repository. Each gate includes line‑level references for precise implementation tracing.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}