Skip to content
SkillModel-led1,926 tokens in use

Explain Interface

Explains how a website, a visual effect or an animation was built, from a live URL or a screenshot, with every claim marked measured, derived or inferred.

12,497 stars on the repository MIT Licensejakubkrehel/skills · skills/explain-interface
npx skills add jakubkrehel/skills
View source

Install command as written in the repository's README

Skill label

measured from its files

What installing Explain Interface costs

26tokens idle
1,926tokens in use
7,712tokens, up to

Idle is what sits in context while the skill is installed. In use is its instructions, loaded when it runs. Up to adds the files its instructions link. Counts are approximate, made offline with cl100k_base.

Model dependence Model-led

Script-ledMixedModel-led

Mostly instructions, so how well it works depends on the model that follows them.

This describes the skill's structure, how much of it is scripts. It does not say how well it works.

Touches

Nothing detected: no shell, network, file writes, git or credentials in its files.

Declared tags come from the skill's frontmatter; detected ones from its scripts, its code blocks and the commands it names in inline code. This is detected, not audited.

How it scores

Quality

How well the skill is written, counted from its own files: its frontmatter, its steps, examples, stated output and constraints, its size and whether every file it ships is one it names.How we score it →
Good
  • Frontmatter8/15
  • Structure17/25
  • Examples6/15
  • Output10/15
  • Size15/15
  • Resources12/15

Security

Starts at 100 and loses points for each thing in the skill's files that a reader should look at first.How we score it →
Excellent

No findings.

Repository health

Pre-computed score based on development activity, maintenance, community, maturity, and trend momentum.How we score it →
Good
  • Development activity96
  • Maintenance52
  • Community48
  • Maturity24
  • Momentum40

Explain Interface is a user-invoked agent skill that answers how something on the web was built. Ask how the gradient on a site was made and it finds the layers behind it in paint order and says what each contributes. Ask how a whole site was built and it reports the frontend stack, tokens, type, spacing, color, motion and breakpoints.

It explains and issues no verdict. Every claim carries a tier, measured, derived or inferred, and a screenshot answer is labeled a reconstruction.

What You Get

  • The layer stack behind an effect in paint order, including pseudo-elements, blur, blend and backdrop layers, with the declaration doing the work on each
  • A site-level read of framework, styling approach, component primitives, tokens, type scale, spacing rhythm, radii, shadows, motion and breakpoints
  • A tier on every value: measured from the page, derived by computation or inferred as intent
  • For screenshots, a reconstruction with exact colors and contrast, ratios in place of pixel sizes and a list of what the image hid
  • A closing recipe in words, plus what is not transferable, such as a licensed typeface or a pre-rendered shadow

Common Use Cases

  • Finding out how a landing-page hero gradient or glow was layered
  • Learning how a site’s reveal animation staggers its text
  • Identifying a site’s framework, Tailwind version and token structure
  • Reconstructing a look from a screenshot when no URL is available

What Makes It Unique

It explains the mechanism and does not dump a readout. It treats compiled output as such: three animations on one element is what a stagger helper compiles to. It also refuses to end in a code block, because anything assembled from compiled output is a lookalike tuned to a viewport and typeface you do not have.

Under The Hood

The skill is a SKILL.md with four reference files for finding effects, reading a whole system, screenshots and working without a browser. It has no scripts, but it gives the agent JavaScript snippets to run in a page through a scriptable browser, such as the chrome-devtools MCP, and shell commands for the no-browser route. That route downloads the page’s HTML and linked stylesheets with curl into a temporary directory and searches them with grep. It reads only the sites you name, treats page text as evidence and never as instructions, never submits forms or signs in, and does not edit your project.

Join founders building with open source

Opinionated takes, migration guides, cost-saving tips, and insights from the open source ecosystem.

Subscribe on SubstackJoin 750+ subscribers