AgentSkillsCN

PostHog Product Analytics

功能采纳率、用户留存率、用户旅程分析

SKILL.md
--- frontmatter
id: posthog-product-analytics
name: PostHog Product Analytics
description: Feature adoption, retention, user journey analysis
role: product
requiredTools:
  - type: mcpService
    serviceType: posthog

PostHog Product Analytics

When analyzing product usage:

  • •Use posthog_list_events (tool name may have a suffix if multiple PostHog servers exist) to measure feature adoption and usage patterns with projectId, event name, and after/before time range.
  • •Use posthog_list_persons and posthog_get_person to understand user segments and retention.
  • •Prefer posthog_list_insights and posthog_get_insight when they answer the question; otherwise use posthog_list_events with clear filters.
  • •Always specify projectId; call posthog_list_projects first if unknown.
  • •For adoption questions, identify the relevant event names and time range, then summarize counts or trends.
  • •When asked about user journeys, combine events and persons to describe paths and drop-off.

Step-by-step instructions

  1. •Resolve projectId: call posthog_list_projects if unknown; identify the relevant event names or insight.
  2. •For adoption: call posthog_list_events with projectId, event filter, and after/before; summarize counts or trend.
  3. •For retention: use posthog_list_persons and posthog_list_events (e.g. distinctId) to compute returning users or cohorts; summarize over time.
  4. •For user journeys: call posthog_list_events in sequence, optionally filter by personId/distinctId; summarize paths and drop-off.
  5. •Prefer posthog_list_insights / posthog_get_insight when they match the question; otherwise build from posthog_list_events with clear filters and cite them.

Examples of inputs and outputs

  • •

    Input: “How many users used the new export feature last week?”
    Output: Count (or trend) for the export-related event(s), projectId, and date range from posthog_list_events; if no such event, say so and suggest event names to check.

  • •

    Input: “Where do users drop off in the onboarding flow?”
    Output: Short sequence of steps with counts or rates and the main drop-off step; cite event names and filters used.

Common edge cases

  • •Unknown projectId: Call posthog_list_projects and ask which one, or use the only/default project.
  • •No matching events: Report zero or “no events found” and suggest verifying event names or time range.
  • •Ambiguous “feature”: List likely event names or ask the user to confirm which event(s) represent the feature.
  • •API/rate limit: Tell the user and suggest a smaller range or retry later.

Tool usage for specific purposes

  • •posthog_list_events: Use for adoption counts, funnels, and journey steps. Always set projectId and after/before time range; use event and distinctId/personId filters as needed.
  • •posthog_list_persons / posthog_get_person: Use for retention and segments.
  • •posthog_list_insights / posthog_get_insight: Use when saved insights answer the question directly.