agentleFS
igovetigovet/cortex-orch-acp/skills/design-reference-analysis/SKILL.md

design-reference-analysis

Skill1 starsChanged 59 days agoMITView on GitHub

Derive transferable design principles (type, spacing, color, composition, density) from reference screenshots of well-made interfaces, recording what to take and what never to copy. Use in the research stage before a design direction is set, when the user supplies references or asks why a reference works.

The file

We split this skill into the rules it gives an agent

4 rules. About 0 are common, 2 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈2

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill to learn from real interfaces the user points to, before a design direction is chosen.

    line 8
  2. Very specific:

    Do not recommend transferring or introducing one-sided accent stripes or decorative leading dots/meta-chips.

    line 54
  3. Distinctive:

    Do not infer a brand's intent, business model or private data from its page.

    line 59
  4. Very specific:

    No self-scoring as evidence: the analysis does not certify that the target will look good.

    line 62

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

Reference analysis

Use this skill to learn from real interfaces the user points to, before a design direction is chosen. The output is a set of principles with their trade-offs, mapped to the target product. It is not a copy of any reference.

This skill is Cortex-authored. It contains no text, markup, assets or values copied from any upstream project (see CORTEX-SOURCE.md).

Inputs

  • Reference screenshots that a writing role captured into the project, or screenshots the user supplied. Read them with the project's image-capable read tool.
  • A URL can be analyzed only if a browser tool is available in this session and it captures the page. If no such tool is available, say that the URL was not observed and ask for screenshots. Never describe a page you have not seen.

Method

  1. Measure. For each screenshot, record what is observable:
    • type: families (by visual classification, marked as estimates), size steps, line height, measure, weight contrast;
    • space: the apparent spacing rhythm and section padding;
    • color: the roles of the main colors (background, text, accent, neutral ramp) with approximate values, marked as estimates unless taken from a supplied file;
    • composition: grid, alignment, density, the proportion of image to text, and the element that carries the first impression;
    • motion: only if a recording or a described sequence was supplied.
  2. Find patterns. Which decisions recur across the references, and which devices carry hierarchy (scale, contrast, whitespace, structure, imagery)?
  3. State principles with trade-offs. Each principle needs: what it does, why it suits this audience, what it costs, and when it fails.
  4. Transfer. Map each principle to the target product and list what must not be transferred: one-to-one layout, logos, copy, imagery, distinctive assets and brand marks.

Output

Return a table with the columns: observation, evidence (screenshot path and region), principle, trade-off, and transfer note. Follow it with a short list of decisions the owner must confirm, and an explicit anti-copy statement: the principles were taken for the target product, and no markup, asset, layout or copy from the references is reused.

Guards

  • Do not recommend transferring or introducing one-sided accent stripes or decorative leading dots/meta-chips; follow the hard bans in the banned AI-tell patterns (read_skill id design-banned-patterns).
  • Take principles and tokens, never markup, assets or one-to-one layouts.
  • Mark every value as measured (from a supplied file) or estimated (from a visual reading). Do not present an estimate as an exact specification.
  • Do not infer a brand's intent, business model or private data from its page.
  • Saving the analysis into project documentation is a write step. A read-only role returns the analysis in its report; a writing role saves it.
  • No self-scoring as evidence: the analysis does not certify that the target will look good. Rendered results are reviewed with design-visual-qa, and quality claims rest on opened screenshots, concrete findings and re-checks.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.