audit
Compare original and translation side by side
🇺🇸
Original
English🇨🇳
Translation
ChineseAudit
审计
Run a technical quality review of the target UI without silently drifting into implementation work.
对目标UI开展技术质量评审，过程中不得擅自开展实现工作。
Preparation
准备工作
- Load first.
impeccable - If no design context exists yet, run before continuing.
impeccable teach - Scope the audit to the named feature, page, screen, or component if an argument was supplied.
- 先加载。
impeccable - 若尚无设计上下文，请先运行，再继续后续操作。
impeccable teach - 若提供了参数，将审计范围限定为指定的功能、页面、屏幕或组件。
What to inspect
检查维度
Score each dimension from 0 to 4:
- Accessibility
- Performance
- Theming and token usage
- Responsive behavior
- Anti-pattern resistance
从0到4分对以下每个维度进行评分：
- 可访问性
- 性能
- 主题与令牌使用
- 响应式表现
- 反模式抗性
Anti-pattern verdict comes first
先给出反模式判定
Start the report by answering one question plainly:
Does this look generic or AI-generated?
Call out specific tells such as gradient text, accent side stripes, purple-blue default palettes, nested cards, or weak typographic hierarchy.
报告开头需直接回答一个问题：
该UI看起来是通用模板或AI生成的吗？
指出具体特征，例如渐变文本、侧边强调条纹、紫蓝色默认调色板、嵌套卡片或薄弱的排版层级。
Report shape
报告格式
- Audit score out of 20
- rating band
- issue counts by severity (through
P0)P3 - top critical findings
- positive findings worth keeping
- recommended next actions
For each issue include:
- location,
- category,
- user impact,
- fix direction,
- and the most relevant follow-up flow (,
polish,typeset,colorize, or a domain-pack implementation pass).adapt
- 审计总分（满分20分）
- 评级区间
- 按严重程度划分的问题数量（至
P0）P3 - 首要关键发现
- 值得保留的正面发现
- 建议后续行动
每个问题需包含：
- 位置，
- 类别，
- 用户影响，
- 修复方向，
- 以及最相关的后续流程（、
polish、typeset、colorize或领域包实现环节）。adapt
Rules
规则
- Be specific, not generic.
- Prioritize ruthlessly.
- Do not fix the issues unless the user explicitly asked for remediation in the same request.
- Prefer verifiable findings over vague taste claims.
- 表述具体，避免泛泛而谈。
- 严格区分优先级。
- 除非用户在同一请求中明确要求修复，否则不得自行修复问题。
- 优先采用可验证的发现，而非模糊的主观评价。