skills/ux-auditor/SKILL.md
Exhaustively audit a top-level UI implementation component against an HTML prototype and produce a grouped markdown checklist of corrections. Use when a user asks for UI parity review, visual QA, design implementation audit, pixel-level drift detection, or behavior/style mismatch analysis between prototype HTML and shipped component code.
npx skillsauth add ryan-mahoney/ryan-llm-skills ux-auditorInstall this skill globally with one command. Works with Claude Code, Cursor, and Windsurf.
3 of 9 scanners reported clean
Some scanners were skipped, did not run, or reported a non-clean status. Review each row below.
Audit an implemented UI against a prototype with element-by-element rigor. Produce a correction-first checklist that is grouped, actionable, and explicit about current vs expected behavior.
Require exactly two arguments in $ARGUMENTS:
Use this format:
$ARGUMENTS="<prototype-html-path> <top-level-component-path>"
If either argument is missing or unreadable, stop and ask for the missing path.
Deviation Taxonomy section in this file as the canonical audit lens.design-system.md before recommending removal or change.design-system.md by checking project root first, then searching with rg --files | rg 'design-system\\.md$'.Design-System-Allowed.Needs-Correction or Needs-Design-Decision with rationale.Correct ....Return markdown using this exact structure:
# UX Deviation Checklist
## Inputs
- Prototype: <path>
- Implementation: <path>
## Audit Summary
- Areas reviewed: <count>
- Elements reviewed: <count>
- Total deviations: <count>
- Severity counts: P0 <n>, P1 <n>, P2 <n>, P3 <n>
- Design-system consultations: <count>
## <Area Name>
### <Category Name>
- [ ] [P1] Correct <specific element/behavior> to match prototype.
Current: <implementation reality>.
Expected: <prototype requirement>.
Evidence: <selector/text/node in prototype> vs <component/style reference>.
Design-system: <Allowed | Not Found | Conflicts | N/A>.
## Design-System Decisions
- <Only include items where design-system affected the recommendation>
## Unclear Items Needing Design Input
- <Only include unresolved uncertainties>
If no deviations are found, still return the same structure and explicitly set Total deviations: 0.
P0: Breaks core task flow or accessibility-critical behavior.P1: Strong visual/interaction mismatch likely noticeable to users.P2: Moderate mismatch in style, spacing, or hierarchy.P3: Minor polish issue with low UX risk.Use these categories to classify every mismatch between prototype and implementation.
When implementation adds elements or styling not visible in prototype:
design-system.md.Design-System-Allowed: addition is supported and non-conflicting.Needs-Correction: addition conflicts with prototype or design-system rules.Needs-Design-Decision: no clear design-system support; flag for product/design judgment.Treat unsupported additions as deviations unless explicitly justified by design-system.md.
documentation
This skill should be used when the user asks to "write a spec", "create a spec", "spec this out", "plan this feature", or "write an implementation plan" for a feature or change. Creates a structured implementation spec in .specs/<slug>/spec.md and mirrors it to GitHub only when the current repository is hosted on GitHub.
data-ai
This skill should be used when the user asks to "execute the spec", "run the plan", "implement the spec", "implement the issue", "run all steps", or "run spec". Implements all steps from .specs/<slug>/spec.md, using a subagent per step when the harness supports subagents.
testing
This skill should be used when the user asks to "review a spec", "review an issue", "check the plan", "review the implementation plan", "find gaps in the spec", or "review spec". Reviews .specs/<slug>/spec.md for gaps and viability, edits it when needed, and mirrors changes to GitHub only when a GitHub mirror exists.
development
This skill should be used when the user asks to "remediate the audit findings", "fix the spec violations", "close the audit findings", "fix conformance violations", or "spec remediate". Reads a spec-audit report, drives one smart subagent per VIOLATION to converge the code back to the frozen spec, and re-audits until clean. Edits production code; never rewrites the spec.