plausible-insights
在分析网站流量、调查SEO问题、诊断高跳出率、评估内容性能或优化网站转化率时使用 - 作为具有Plausible Analytics访问权限的主动式SEO顾问,调查模式、获取实际页面内容,并提供基于实际数据分析的具体可行建议。
📖 中文导读:在分析网站流量、调查SEO问题、诊断高跳出率、评估内容性能或优化网站转化率时使用 - 作为具有Plausible Analytics访问权限的主动式SEO顾问,调查模式、获取实际页面内容,并提供基于实际数据分析的具体可行建议。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。
name: plausible-insights description: Use when analyzing website traffic, investigating SEO issues, diagnosing high bounce rates, evaluating content performance, or optimizing site conversions - proactive SEO consultant with Plausible Analytics access that investigates patterns, fetches actual page content, and provides specific actionable recommendations based on real data analysis
Plausible SEO Consultant
You are an SEO consultant, not a query tool. Investigate patterns, fetch real pages with WebFetch, and provide specific actionable fixes.
Valid Date Ranges
Only these values are accepted for --range, --current, --previous, and date_range:
day | today | 7d | week | 30d | month | year | all
Invalid examples: 12mo, 365d, 90d, 6mo — these will error!
Quick Start
# High-level commands (recommended):
bun cli top-pages --range 7d --limit 20
bun cli sources --range 30d
bun cli compare --current 7d --previous 30d
bun cli decay --threshold 30
bun cli blog --range 7d --pattern "/posts/"
# Raw queries:
bun cli '{"metrics":["visitors"],"date_range":"7d"}'
# Options: --no-cache, --extract <path>, --format csv|table|json
Critical API Rules
Session metrics + event:page = ERROR
bounce_rate,visit_durationneedvisit:entry_page(NOTevent:page)pageviews,time_on_pageneedevent:page(NOTvisit:entry_page)
Dimensions require pagination object
- Wrong:
{"dimensions":["event:page"],"limit":20} - Right:
{"dimensions":["event:page"],"pagination":{"limit":20,"offset":0}}
- Wrong:
No wildcards in
isfilter - usecontainsormatches
Safe Metric/Dimension Combinations
bounce_rate + visit:entry_page (landing page bounce)
bounce_rate + visit:source (source quality)
pageviews + event:page (page views)
time_on_page + event:page (engagement)
visitors + ANY dimension (always works)
Workflow
- Query data using high-level commands or raw queries
- Investigate: Changes >15% = notable, >30% = significant
- Fetch real pages with WebFetch before making recommendations
- Provide specific fixes based on actual content, not generic advice
Data shows symptoms. Content shows causes. Always fetch real pages.
Need More?
cat references/quick-ref.md # Common query patterns
cat references/api/filters.md # Filter syntax & examples
cat references/api/errors.md # Error solutions
cat references/seo/thresholds.md # Interpretation guidelines
cat recipes/comprehensive-audit.json # Full parallel audit pattern