Does Hallmark Clone Pixels with the Study Verb?

The hallmark study verb is expressly designed not to copy pixels from a reference design, instead extracting only the structural "design DNA" to enable original rebuilds.

The study verb in the Hallmark CLI (found in the Nutlope/hallmark repository) handles design analysis through a strict protocol that prioritizes architectural understanding over visual duplication. When you point the tool at a screenshot or URL, it runs a diagnostic extraction pipeline that captures macrostructure, typographic pairing, and color anchors while explicitly refusing to replicate the original bitmap.

The Core Anti-Clone Contract

Hallmark’s pixel-refusal policy is codified across three key documentation files that define the verb’s behavior.

In skills/hallmark/SKILL.md, the high-level definition states: "Never copies pixels. Refuses pixel-clones and paid templates." This establishes the foundational constraint that separates Hallmark from screenshot-to-code tools.

The detailed protocol in skills/hallmark/references/study.md reinforces this boundary explicitly: "It does not copy pixels. It does not output a façade of the source." This document governs the extraction logic and refusal mechanisms when users request literal reproductions.

User-facing documentation in docs/study-examples.md clarifies that the verb "extracts structural DNA … never pixels," noting that achieving a 1:1 pixel clone requires a different category of tool entirely.

The Five-Step Extraction Pipeline

When processing a screenshot or URL, Hallmark executes a diagnostic sequence that translates visual input into architectural metadata:

1. Surface – Identifies base canvas constraints and viewport characteristics
2. Type – Extracts font pairings, weights, and hierarchy relationships
3. Structure – Maps component archetypes and macrostructure (e.g., "Marquee Hero" layouts)
4. Motion – Detects animation patterns and transition timing
5. Rhythm – (In image mode) analyzes spacing intervals and compositional balance

The output is a markdown diagnosis report containing attributes like macrostructure = "Marquee Hero", colour anchor = "warm-red", and type-pairing = "italic editorial serif + neutral grotesque body". This structured data provides the blueprint for rebuilding the design with original assets rather than cloned pixels.

Using the Hallmark Study Command

The study command accepts both local image files and live URLs, initiating the extraction pipeline without generating CSS or HTML that mirrors the source pixels.

To diagnose a design from a local screenshot:

hallmark study my-inspiration.png

To analyze a live webpage:

hallmark study https://example.com/awesome-site

Both commands return a markdown diagnosis describing the extracted design DNA. The tool does not emit code that reproduces the original visual execution; it only prepares the structural skeleton for a subsequent rebuild using the hallmark build or similar generation commands.

Refusal Logic for Pixel Clone Requests

If a user explicitly requests a literal pixel-by-pixel copy, Hallmark activates refusal protocols defined in study.md. The tool will decline the operation and suggest alternative approaches, maintaining the architectural focus over visual replication. This safeguards against copyright infringement and ensures outputs remain original implementations inspired by—but not duplicating—reference materials.

Summary

• Hallmark’s study verb extracts design DNA (macrostructure, typography, color anchors) while explicitly refusing to clone pixels.
• The "Never copies pixels" rule is documented in SKILL.md, study.md, and study-examples.md across the repository.
• The five-step pipeline (Surface → Type → Structure → Motion → Rhythm) outputs diagnostic metadata, not bitmap reproductions.
• Commands like hallmark study my-inspiration.png generate markdown reports describing structural attributes, not CSS or HTML facades.
• Refusal mechanisms prevent pixel-cloning requests, redirecting users toward proper rebuild workflows.

Frequently Asked Questions

Does Hallmark output CSS that matches the reference pixel-for-pixel?

No. The study verb outputs a markdown diagnosis describing structural attributes like component archetypes and type pairings. It does not generate CSS, HTML, or any code that reproduces the original bitmap. If you need pixel-perfect replication, Hallmark will refuse and recommend alternative tools.

What files define the study verb's refusal to copy pixels?

The constraint is documented in three locations: skills/hallmark/SKILL.md establishes the "Never copies pixels" rule, skills/hallmark/references/study.md details the extraction protocol and refusal logic, and docs/study-examples.md explains the structural DNA approach to end users.

Can I use Hallmark study on both images and live URLs?

Yes. The command accepts both local image files (e.g., hallmark study screenshot.png) and live URLs (e.g., hallmark study https://example.com). Both inputs trigger the same five-step extraction pipeline that analyzes surface, type, structure, motion, and rhythm without cloning the original pixels.

What happens if I ask Hallmark to recreate a design exactly as shown?

Hallmark will refuse the request. According to the source code documentation in study.md, the tool "does not output a façade of the source" and will decline literal pixel-copy operations. Instead, it provides the structural diagnosis necessary to rebuild the design using original visual execution.

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