improve-ui

Audit product interfaces against design evidence and write implementation plans for verified UI problems.

Updated Mar 12, 2026
One-click install
npx skills add https://github.com/RavitejaKarra24/dotfiles --skill improve-ui-ravitejakarra24
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: improve-ui
Source: https://github.com/RavitejaKarra24/dotfiles/tree/main/agents/.agents/skills/improve-ui
Command: npx skills add https://github.com/RavitejaKarra24/dotfiles --skill improve-ui-ravitejakarra24

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes references (resource) components.

What problem does it solve? UI reviews often produce vague, unsupported suggestions or accidental source changes. This Skill audits a product surface strictly against its own governing design system, proves every finding with contract, runtime, and correction evidence, and produces self-contained implementation plans without ever modifying product source code. ## Core Features & Use Cases - Evidence-gated UI audit: Traces the rendered path of one coherent surface, reconstructs its local design system, and keeps only findings backed by a binding design contract, a proven runtime path, and a single deterministic correction. - Read-only safety: Never edits product source, installs dependencies, or commits; all output lives under design-plans/ as Markdown plans another agent can execute. - Structured reporting and handoff: Reports at most three vetted findings in a confidence-ranked table, recommends the highest-leverage improvement, then writes self-contained plans using the included plan template. - Use Case: Ask it to review your app's settings page; it traces the components and tokens that actually govern that surface, reports only contradictions it can prove against documented design decisions, and drafts an implementation-ready plan for the finding you select. ## Quick Start Use the improve-ui skill to audit the primary product surface, identify supported UI problems, and recommend what to improve first without modifying anything.

Frequently Asked Questions about improve-ui

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I audit a UI against its design system with an AI agent?

Invoke the skill with a target surface, and it traces the rendered path through routes, layouts, components, and tokens, then reconstructs the governing design system from current documentation. Only findings with a binding contract, proven runtime path, and single correction survive.

What counts as a valid UI finding in a design audit?

A finding must correct visual presentation, copy, layout, styling, or a documented design rule, backed by three proofs: a binding design contract, a traced runtime path to the surface, and one deterministic correction. Functional bugs, accessibility issues, and code-quality concerns are excluded unless explicitly requested.

Does the improve-ui skill modify my source code?

No, it is strictly read-only on product source. It never edits code, installs dependencies, runs formatters, or commits; it only creates or updates Markdown plan files under the design-plans/ directory for another agent to execute.

Can it audit an entire repository at once?

No, it selects one deployable application and one coherent surface family representing a primary product task. Repository-wide searches produce candidates, not findings, and unrelated applications, previews, and legacy systems are excluded from the traced path.

What happens when no UI problems are found?

If no candidate survives the proof and falsification passes, the report states that no supported findings were found and gives no recommendation. The skill prefers reporting nothing over returning an unsupported finding.