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/agricidaniel/codex-seo --skill seo-visual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。
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。
Visual Specialist
Use this for the visual sub-track in full audits.
Shared Data Cache
Step 0 -- Check shared data cache:
Before gathering, check .seo-cache/ for cached data from other skills. Reference: ../seo/references/shared-data-cache.md for schemas and dependency map.
This specialist does not require upstream cache inputs before running fresh visual checks.
- If found: no prerequisite cache is needed; continue with current visual capture workflow
- If missing: proceed normally and gather fresh visual data
- If JSON is corrupt or unreadable: treat it as missing and continue normally
- If the user says "refresh" or "re-run": ignore cache entirely
Inputs
- URL
- Timeout
- Visual mode (
on|off|auto)
Outputs
VISUAL-AUDIT-REPORT.mdSUMMARY.jsonscreenshots/(if Playwright available)
Checks
- H1 and CTA visibility above the fold
- Mobile viewport + horizontal scroll
- Touch target sizing and minimum font size
- Multi-viewport screenshots
Write to shared data cache
After completing all work, write results to .seo-cache/.
mkdir -p .seo-cache/pages/{url-slug}Write .seo-cache/pages/{url-slug}/visual.json using the schema in ../seo/references/shared-data-cache.md. Preserve the documented top-level keys: cache_type, analyzed_at, url, url_slug, score, layout_summary, and issues. Add extra visual detail only as supplemental fields, such as above_fold, mobile, screenshots, or notes.
Add .seo-cache/ to .gitignore if not already present:
grep -qxF '.seo-cache/' .gitignore 2>/dev/null || echo '.seo-cache/' >> .gitignorePremium Deliverable
If the user requests a 'client report' or 'premium deliverable', automatically read ../seo-audit/assets/report-template.html and ../seo/references/premium-report-standard.md. Use that report standard as the default brief. Generate the HTML as an internal intermediate artifact, then produce the PDF as the public deliverable. Mention only the PDF in your final response unless the user explicitly asks for the HTML file.
