Heuristic audit
Audit an existing interface for usability failures, operational impact, and minimum defensible fixes.
Canonical skill profile. This page is generated from the repository contract. The linked
SKILL.mdremains the source of authority.
| Category | Version | Owner |
|---|---|---|
design |
0.1.0 |
adaptive-skills |
Jump to
- Overview
- When to use
- When not to use
- Core moves
- Expected output
- Verification
- Handoff signals
- Pairs well with
- Anti-patterns
View the canonical contract on GitHub.
Overview
Use this skill when the interface exists and you need a practical, prioritized read of what is breaking down.
When to Use
- screen reviews
- existing flows with friction
- dashboard and workspace audits
When NOT to Use
- choosing a new strategic direction before the interface exists
Core Moves
- Choose the review depth.
- Identify a small set of meaningful findings.
- Name the heuristic behind each issue.
- Describe the operational impact.
- Suggest the minimum defensible correction.
Optional Modules
- Severity pass — Sort issues by operational harm rather than taste.
- State coverage — Check empty, loading, and failure states when the problem spans system feedback.
- Accessibility lens — Add focus, contrast, semantics, or assistive-tech checks when relevant.
Activation Triggers
- Use severity pass when the list of issues is long.
- Use state coverage when dynamic states are central to the experience.
- Use accessibility lens when the interface is dense or high-risk.
Expected Output
- prioritized findings
- heuristic mapping
- minimum fixes
Verification
- Findings are concrete and observable.
- Each issue explains why it matters operationally.
- Recommendations do not balloon into a redesign without cause.
Handoff Signals
- The next step is visual redesign or content work.
- A product or technical decision must change before UX can improve.
Pairs Well With
ux-writingux-strategy
Anti-patterns
- Using heuristics as a decorative checklist.
- Reporting style preferences as usability failures.
- Listing too many low-value findings.