Hallmark

Description

Hallmark is an opinionated design skill for AI coding assistants. It encodes a tight set of rules drawn from the anti-AI-slop design field and refuses to let the model fall back to the defaults every LLM was trained on. Its differentiator is structural variety, not just visual variety: two pages built from two different briefs should not share the same hero → 3-feature → CTA → footer rhythm. One default behaviour and three verbs — audit (score a target against the anti-pattern list, no edits), redesign (new visual structure inside the existing implementation boundaries), and study (extract the DNA from a screenshot or URL). Ships 21 named themes plus a custom OKLCH palette branch, a 50+ gate slop test, and a pre-emit self-critique that scores every output on six axes before handing it back. Redistribution of the MIT-licensed upstream project. See NOTICE.md and LICENSE.

What you get

• 21 named themes with a diversification rule, plus a custom OKLCH palette branch for briefs that need one
• A 50+ gate slop test run before every emit, with a six-axis self-critique stamped into the artifact
• Mobile verified at 320 / 375 / 414 / 768 px as a hard floor, not a wish list
• Implementation safety rails: no deletions without an explicit file-level plan

Supported agents

Usage and examples

1. Drop the unpacked folder into your agent's skills directory
2. Ask for a page or app as usual — the design flow runs by default
3. Run "hallmark audit <target>" to score an existing page without editing it
4. Run "hallmark redesign <target>" to replace the visual layer in place
5. Run "hallmark study <screenshot or URL>" to extract a reference's DNA

Requirements and limitations

Onchain Identity

Public Metadata

Files