使用方法
- 复制安装命令，在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物，并对照来源检查输出。
npx skills add https://github.com/sergekostenchuk/ui-ux-agent-skill-system --skill ux-audit-skill开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况，不等于独立质量或安全测试。
仓库原始内容
按源仓库内容呈现，保留标题、案例、代码、表格、链接以及原文引用的演示图片。
UX Audit Skill
Use this specialist for evidence-backed review of existing UI artifacts. The output is a severity matrix and actionable refactor brief, not a generic design essay.
Operating Modes
review-live-url: audit a reachable URL with available browser/static evidence.review-screenshot: audit image-only context without DOM claims.review-code: audit source files and templates.refactor-brief: convert findings into implementation tasks for another skill.regression-check: compare before/after evidence and remaining risks.
Required Workflow
- Classify available evidence using references/audit-evidence-model.md.
- Apply references/severity-model.md before ranking findings.
- Use assets/audit-report.template.md for audit output.
- Use assets/regression-report.template.md for before/after checks.
- Apply
../../shared/privacy-policy.mdbefore screenshots, Figma, or authenticated state. - If a URL is available, reuse local scripts listed in
../../shared/tool_inventory.json. - Label every finding by evidence type:
DOM,screenshot,source,automated, orManual.
Evidence Boundaries
- Screenshot-only input cannot support DOM, Lighthouse, axe, or runtime claims.
- Source-only input cannot prove rendered layout unless run or screenshot evidence exists.
- Manual findings are allowed but must be labeled
Manual. - Planned tools are not evidence.
Validation
bash
python3 $CODEX_SKILLS_DIR/.system/skill-creator/scripts/quick_validate.py $CODEX_SKILLS_DIR/ux-audit-skill
python3 $CODEX_SKILLS_DIR/senior-skill-architect/scripts/lint_production_skill.py $CODEX_SKILLS_DIR/ux-audit-skillOutput Shape
text
Audit target:
Evidence available:
Ran:
Skipped:
Severity matrix:
Top fixes:
Verification plan:
Privacy notes:
Residual risk:
