SuperAPI AI Directory

review-harmonyos-design

Design & creation

View repository

Updated 10-15 21:12

About this skill

Rigorously Review HarmonyOS Design

This Skill only performs review; it does not design from scratch, modify code in bulk, or review engineering issues unrelated to interface design.

Default stance:Passing requires evidence, and problems also require evidence.

1. Review Scope

  • User tasks and navigation;
  • Adaptation for phones, tablets, foldables, PCs, wearables, smart screens, and car machines;
  • Input from touch, mouse, keyboard, remote control, etc.;
  • Normal, disabled, pressed, focused, activated, and hovered states of controls;
  • Visual Tokens, vp/fp, large fonts;
  • Motion purpose, curve, duration, follow-finger, speed, and interruptibility;
  • Truthfulness of asynchronous states;
  • Accessibility;
  • Animation performance;
  • Stable principles and versioned platform layers;
  • Evidence from interactive prototypes and real implementations.

2. Out of Scope

For the following requests, state that they are not applicable and point to general engineering review:

  • HAP/APP signing;
  • Compiler errors;
  • Networking, permissions, databases;
  • Non-UI business logic;
  • Design specifications of other mobile platforms;
  • Directly implementing new features.

3. Establish Context First

Reuse information already provided by the user:

目标设备:
窗口形态:
主要输入:
API_Level:
主题与字号:
任务路径:
证据:
假设:

When missing information does not affect the current finding, do not ask repeatedly. Conclusions with insufficient evidence should be marked as low confidence.

4. Thirteen Non-Negotiable Standards

  1. Clear navigation. Users know where they are, where they are going, how to go back, and how to exit.
  2. Cross-device is not proportional scaling. There must be a specific adaptation strategy.
  3. System first. Customization must not weaken system states and accessibility.
  4. Complete input states. Under the target input, pressed, focused, hovered, activated, etc. must be distinguishable.
  5. Immediate feedback. There is local feedback during the press phase.
  6. Continuous gestures. Follow-finger, release velocity inheritance, and target changes must be interruptible.
  7. Motion expresses real relationships. Curves, directions, shared elements, and hierarchy are consistent.
  8. Asynchronous states do not lie. Pending must not masquerade as Confirmed; success feedback must not precede real confirmation.
  9. Separation of foundations and versions. The current visual language must not overturn direction sense, readability, accessibility, and state truthfulness; old principles must not block new platform capabilities either.
  10. Product context first. Do not turn all products into the same kind of glass, rounded corners, bounce, or "premium feel".
  11. Executable prototype. High-risk gestures, transitions, and cross-device solutions must not pass based on static mockups alone.
  12. Accessible. Non-text semantics, states, focus, large fonts, and custom-drawn content must be usable.
  13. Performance has evidence. Avoid hand-written per-frame animation, continuous layout animation, and redundant refreshes, and validate on real devices.

Detailed review criteria are set out in references/STANDARDS.md。

5. Issues that warrant immediate escalation

Typically marked as BLOCKER or MAJOR:

  • A key page has no way to go back or exit;
  • core actions are clipped on tablets or with large fonts;
  • core buttons have no pressed feedback;
  • the velocity drops to zero after a gesture is released, causing a noticeable pause;
  • input is locked during animation;
  • shared transitions and the default Navigation transition are stacked;
  • success is displayed or announced before a remote request is confirmed;
  • icon buttons have no accessible name;
  • focus traps;
  • high-frequency gestures modify width/height;
  • House Style is claimed to be official rules;
  • current visual trends are used as a reason to remove necessary contrast, focus, or directionality;
  • high-risk gestures have only static comps, with no runnable prototype or target-device evidence.

6. Fix priority

  1. Fix tasks, navigation, and state truthfulness;
  2. restore system default capabilities;
  3. fix cross-device and input;
  4. fix accessibility;
  5. remove purposeless motion;
  6. fix tracking, velocity, and interruption;
  7. unify semantic tokens;
  8. optimize performance;
  9. finally handle brand micro-motion.

7. Required output format

Context card

List known conditions and assumptions.

Findings table

Priority Rule ID Evidence Issue User impact Recommendation ArkUI touchpoint Source level Confidence

One issue per row. Issues without evidence do not go into Findings.

Impact summary

In the following order, output only non-empty categories:

  1. Tasks and navigation;
  2. asynchronous truthfulness;
  3. cross-device and input;
  4. gestures and motion;
  5. accessibility;
  6. performance;
  7. stability principles and platform versions;
  8. visual consistency;
  9. prototypes and evidence;
  10. to be verified.

Minimal fix plan

Give dependency order, scope of changes, and verification methods; do not directly rewrite the entire project.

Conclusion

Can only be:

  • Fail: Blocker exists;
  • Conditional pass: no Blocker, but unresolved Major issues exist;
  • Pass: no Blocker/Major, and evidence exists for key devices, input, font sizes, accessibility, and performance.

8. Source wording

  • H1: Huawei official;
  • H2: OpenHarmony official;
  • H3: official observation;
  • H4: project recommendation.

Do not write 0.97, unified springs, staggered entrances, glass materials, etc. as official HarmonyOS requirements.

9. Anti-homogenization check

For each review, add:

应保留的产品特征:
不应套用的通用风格:
允许的 House Style:

The goal of the review is appropriateness, not trendiness.