Back to directory

Hallmark

A design skill for AI coding agents such as Claude Code, Cursor and Codex that builds web page UI while avoiding common AI-generated design defaults, using a set of automated checks.

Hallmark

Hallmark is a design skill for Claude Code, Cursor and Codex that generates web page UI while avoiding common AI-generated design defaults. For a given brief, it picks a page macrostructure and one of twenty-one visual themes, then runs fifty-seven checks plus a self-critique pass before returning the result, aiming to make two different briefs produce visually distinct pages rather than colour variations of one template.

It offers four commands: a default build command; an audit command that scores existing code against the same anti-patterns without editing it; a redesign command that keeps a page's copy, information architecture and brand while rebuilding its structure; and a study command that extracts the macrostructure, type pairing and colour anchor from an admired design, optionally producing a portable design document. A custom mode designs a page from scratch when no catalog theme fits the brief.

Each generated page is self-contained HTML and CSS. It installs as a skill through the skills command-line tool or by copying its files into a Claude Code or Cursor skills folder, and is licensed under the MIT license.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.