// the find
Nutlope/hallmark
Anti-AI-slop design skill for Claude Code, Cursor, and Codex.
Hallmark is a prompt-engineering skill (a structured SKILL.md + reference files) for AI coding assistants that tries to stop them from generating cookie-cutter landing pages. You drop it into Claude Code, Cursor, or Codex and it guides the model to pick from 21 macrostructures, apply one of 20 themes, and run 57 anti-slop checks before emitting HTML. The target user is a developer who needs to ship a decent-looking marketing page without a designer.
The 57-gate slop test is the real differentiator — it's a concrete, enumerable checklist that gives the LLM something to falsify rather than just a vague 'make it look good' instruction. The macrostructure library (21 named layouts with individual spec files) is genuinely useful design vocabulary; naming 'marquee-hero' vs 'stat-led' vs 'letter' forces structural variety instead of hero→features→CTA by default. The study verb that extracts design DNA from a screenshot into a portable design.md is a clever workflow: it separates the analysis from the generation and lets you hand off to a different tool or session. The example outputs in site/examples are honest proof-of-work — they look visually distinct from each other, which is the whole claim.
This is entirely prompt engineering — the quality of output degrades hard with the model and context window. There's no programmatic enforcement; the 57 slop gates are instructions the LLM grades itself on, so a distracted or cheap model will just say it passed. The skill is CSS-heavy HTML output with no component framework, so if you're building on React/Vue/Svelte you're getting a design reference at best, not something you can drop in. The 20 named themes have only a handful with actual reference files (carnival, cobalt, hum, lumen); most are implicit, which means the model is improvising within a label, not following a real spec.