PlugWiki

better-interface skill: review a whole interface in one ranked report

Makes Claude review a screen, flow or whole interface across six topics and return one ranked report.

Installs
32.5K
on skills.sh
Stars
10K
on GitHub
Files
2
besides SKILL.md
Updated
Oct 2026
last repo commit

better-interface is the review conductor in Jakub Krehel's interface skills. Claude finds the scope, reads your project's own interface docs, then passes the interface through six topic skills: accessibility, layout, writing, typography, color and UI polish. It merges their evidence into one table ranked by severity.

Every finding must cite a file and line, and the report ends with Block or Approve. It reviews only and edits nothing unless you ask for the fixes. It needs the six topic skills installed, so the install box adds them.

Good for

  • Front-end developers who want one pre-launch UI audit
  • Design engineers checking a flow against accessibility and polish rules
  • Teams that want review findings tied to files and lines

Install better-interface

Claude Code

  1. 1

    Copy this prompt.

    Install the better-interface skill in Claude Code. Run this command:
    
    npx skills add jakubkrehel/skills --skill better-interface --skill better-accessibility --skill better-layout --skill better-writing --skill better-typography --skill better-colors --skill better-ui
    
    Then check that it loaded and tell me how to use it.
  2. 2

    Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

better-interface sends the interface through six topic skills, so the first route installs all seven. The plugin route installs all 14 skills in the repo and its manifest lists no hooks or MCP servers. Pick one route, not both. The repo also has interface-review for changes, which you start yourself.

From the vendor's docs. Source.

What's in the skill

When it loads

Reviews a screen, flow or whole interface across accessibility, layout, writing, typography, color and UI polish, and returns one ranked report.

Sections in SKILL.md

  1. Evidence, not taste
  2. 1. Resolve the scope first
  3. 2. Send a change to interface-review
  4. 3. Recon before judgment
  5. 4. Use domain skills as the sources of truth
  6. 5. Verify what can be verified
  7. Require evidence
  8. Rank by user impact
  9. Prefer the cheaper fix
  10. Consolidate systemic findings
  11. Review without mutating by default
  12. Before you finish
  13. Review output format

Key facts

Installs on skills.sh
32,534 as of Nov 2, 2026
License
MIT
Last repo commit
Oct 25, 2026

What it includes

  • Works out the screen, flow or repo to review and states it

    Too big? It narrows to one complete flow and says what it left out.

    Try: “Review the checkout flow in this app.”

    Read · Resolve the scope first

  • Branch, PR or uncommitted work is a change review for another skill

    It asks you to run interface-review, which only you can start, and never guesses a diff.

    Send a change to interface-review

  • Finds the framework, styling, tokens and the project's own interface docs

    Reads CONTRIBUTING.md, CODING_STANDARDS.md, AGENTS.md, CLAUDE.md, design-system docs and ADRs if present.

    Read · Recon before judgment

  • Loads six skills in order and takes their rules and checks

    Order: accessibility, layout, writing, typography, colors, ui. A missing skill is marked Not reviewed.

    Use domain skills as sources of truth

  • Runs safe checks the project offers and inspects the rendered interface

    A check it cannot run is Not verified, never a finding.

    Runs code · Verify what can be verified

  • Every finding cites path/to/file:line and shows the current code

    Require evidence

  • HIGH, MEDIUM and LOW, with 13 escalation triggers that are HIGH on sight

    Examples: no accessible name, no visible focus ring, unreachable by keyboard, fails contrast.

    Rank by user impact

  • Delete, use the platform, reuse what exists, correct the value, then add

    Prefer the cheaper fix

  • One root cause is one row, listing every location; at most 15 findings

    Consolidate systemic findings

  • Read-only review; edits source only if you also ask for the fixes

    Review without mutating by default

  • Scope and coverage, one findings table, verification, then Block or Approve

    Block means a HIGH finding remains. Approve covers only the domains it inspected.

    review-format.md

From the official docs

Permissions

Low risk

It reads your project and may run checks the project already has. Reviews do not edit code unless you ask for the fixes.

What it can do

Watch out for

  • It runs your project's own checks

    Step 5 runs relevant checks the project offers and may inspect the rendered interface. Read each command before you approve it. github.com

  • Asking for fixes lets it edit code

    Reviews do not change source. If you ask it to implement the findings it will, and it re-runs verification afterwards. Work in a branch. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details
Turning on read-onlyA review is read-only by default. Source is edited only if you also ask Claude to implement the findings.
Sign-in and keysSKILL.md names no credentials or sign-in.
Scopes
  • Claude reads your project's source and interface docs

  • Claude may run safe checks your project offers and inspect the rendered interface

Questions

What does better-interface do?

It makes Claude review a screen, a flow or a whole interface for accessibility, layout, writing, typography, color and UI polish, and return one ranked report. Every finding cites a file and line, and the report ends with Block or Approve.

How do I install better-interface?

Use the install box at the top of this page. It installs better-interface with the six topic skills it calls. If you take the plugin route instead, you get all 14 skills in the repo, and the skill is called /interfaces:better-interface.

What happens if a topic skill is missing?

The skill marks that domain Not reviewed, names the missing skill and continues with the rest. It does not recreate that skill's rules from memory or claim full coverage.

Will it edit my code?

Not by default. SKILL.md treats a review request as read-only and edits source only if you also ask Claude to implement the findings.

better-interface vs interface-review?

better-interface reviews a screen, a flow or the whole interface. Reviews of a branch, pull request or uncommitted changes belong to interface-review. That skill is started by you, so better-interface asks you to run it instead of starting it.

better-interface vs better-accessibility?

better-accessibility is one of the six topic skills better-interface calls. Install it alone for a WCAG 2.2 review of keyboard, focus, forms and screen readers. Use better-interface when you want all six topics in one ranked report.

Does it work in the Claude app?

The README documents the skills CLI and the Claude Code plugin only. The install box leaves out a Claude app tab because the README documents no upload route for this set of seven skills.

Is better-interface safe?

It reads your project and may run the safe checks the project already offers, and it may inspect the rendered interface. It bundles no scripts of its own. Read each command before you approve it.

Information

updated Nov 2, 2026

Folder
skills/better-interface

jakubkrehel/skillsskills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 7
  1. jakubkrehel/skills README · github.com
  2. skills CLI commands · github.com
  3. better-interface SKILL.md · github.com
  4. plugin manifest · github.com
  5. AGENTS.md · github.com
  6. review-format.md · github.com
  7. skills.sh listing · skills.sh

Updated Nov 2, 2026.