Nielsen UI Review
name: nielsen-ui-review
by dawen-ye · published 2026-03-22
$ claw add gh:dawen-ye/dawen-ye-ui-review---
name: nielsen-ui-review
description: >-
Run a structured UI/UX heuristic review based on Nielsen's 10 usability principles for screenshots, mockups, wireframes, product pages, app screens, and design comps. Use when the user provides a UI image or screen and asks to review, critique, evaluate, audit, or assess the interface, usability, UX, or design quality, including requests such as "review this UI", "evaluate this interface", "do a heuristic evaluation", "give me UI feedback", "assess usability", or Chinese requests like "评价一下这个界面", "帮我做 UI 评审", or "做个可用性分析". Ask a short clarification round first, using selectable options when possible, then generate prioritized findings and actionable recommendations.
---
# Nielsen UI Review
Review a UI image through the lens of usability, not personal visual taste. Gather minimal context first, then produce a structured review based on Nielsen's 10 heuristics.
Core Behavior
Workflow
1. Inspect the screenshot.
2. Infer the likely screen type, platform, and primary task.
3. Ask 3 to 5 concise clarification questions.
4. Wait for answers if the user responds.
5. If the user does not respond, proceed with explicit assumptions.
6. Review the screen against Nielsen's 10 heuristics.
7. Summarize the most important issues first.
8. End with prioritized, actionable recommendations.
Inspect First
Identify, if possible:
If the screenshot is blurry, cropped, or too partial for high-confidence review, say so before continuing.
Clarification Round
Ask 3 to 5 questions. Keep them short. Use options when possible. End with:
`If you're not sure, I can still do a first-pass review based on the screenshot and make my assumptions explicit.`
Default Question Bank
Pick the most relevant questions from this list:
1. Product goal
2. Target user
3. Primary task on this screen
4. Business priority
5. Usage context
6. Constraints
Image-Driven Questions
Add 1 or 2 screenshot-specific questions only if they materially affect the review. Example questions:
If No Clarification Arrives
Continue with a short assumption block such as:
`This review is based on the visible screenshot only. I am assuming this is a primary task screen, the visible copy is intentional, and the goal is to help users complete the main action with low confusion.`
Review Standard
Judge usability, not taste.
Do:
Do not:
Nielsen Heuristics
Review the screen against these heuristics:
1. Visibility of system status
2. Match between system and real world
3. User control and freedom
4. Consistency and standards
5. Error prevention
6. Recognition rather than recall
7. Flexibility and efficiency of use
8. Aesthetic and minimalist design
9. Help users recognize, diagnose, and recover from errors
10. Help and documentation
Severity
Assign one severity per issue:
Optionally add confidence:
Use confidence only when the screenshot is ambiguous.
Output Format
Use this exact structure unless the user asks for another format.
Context Summary
Summarize:
Assumptions
List the assumptions used to proceed.
Top Findings
List the 3 to 5 most important findings first. For each finding, include:
Heuristic Review
Cover all 10 heuristics briefly.
For heuristics with a real issue or strength, include:
If no clear issue is visible, write:
`No obvious issue detected from the screenshot.`
Priority Actions
End with:
1. Fix first
2. Improve next
3. Nice-to-have refinements
Make each action concrete.
Open Questions
List any remaining questions that would meaningfully change the review.
Response Style
Keep the tone:
Prefer statements like:
Avoid statements like:
Default Opening Prompt
Use a short opening like this:
I'll start with a usability review based on the screenshot. To make the feedback more relevant, please pick the closest options below.
1. Product goal
2. Target user
3. Primary task on this screen
4. Business priority
If you're not sure, I can still do a first-pass review based on the screenshot and make my assumptions explicit.
Final Check
Before sending the review, verify:
More tools from the same signal band
Order food/drinks (点餐) on an Android device paired as an OpenClaw node. Uses in-app menu and cart; add goods, view cart, submit order (demo, no real payment).
Sign plugins, rotate agent credentials without losing identity, and publicly attest to plugin behavior with verifiable claims and authenticated transfers.
The philosophical layer for AI agents. Maps behavior to Spinoza's 48 affects, calculates persistence scores, and generates geometric self-reports. Give your...