AgentSkillsCN

plausible-insights

在分析网站流量、调查 SEO 问题、诊断高跳出率、评估内容表现或优化站点转化时使用——主动 SEO 顾问,拥有 Plausible Analytics 访问权限,调查模式、获取实际页面内容并提供基于真实数据分析的具体行动建议

SKILL.md
--- frontmatter
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:

code
day | today | 7d | week | 30d | month | year | all

Invalid examples: 12mo, 365d, 90d, 6mo — these will error!

Quick Start

bash
# 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

  1. Session metrics + event:page = ERROR

    • bounce_rate, visit_duration need visit:entry_page (NOT event:page)
    • pageviews, time_on_page need event:page (NOT visit:entry_page)
  2. Dimensions require pagination object

    • Wrong: {"dimensions":["event:page"],"limit":20}
    • Right: {"dimensions":["event:page"],"pagination":{"limit":20,"offset":0}}
  3. No wildcards in is filter - use contains or matches

Safe Metric/Dimension Combinations

code
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

  1. Query data using high-level commands or raw queries
  2. Investigate: Changes >15% = notable, >30% = significant
  3. Fetch real pages with WebFetch before making recommendations
  4. Provide specific fixes based on actual content, not generic advice

Data shows symptoms. Content shows causes. Always fetch real pages.

Need More?

bash
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