posthog-foss: Skill for Claude Code

Analyzing insights across teams

system.* entity tables (e.g. system.insights) are scoped to the current project, and the generic execute-sql guidance says other teams' data is inaccessible. For the dogfood project (US project 2) that is not the whole story: production Postgres tables are replicated into the project's data warehouse, so cross-team entity metadata is queryable with posthog:execute-sql. Do not stop at system.insights when the question spans teams.

This skill is deliberately repo-local (.agents/skills/): it documents PostHog's internal dogfood setup, applies only to agents working in this repo, and must not move into the packaged products/*/skills/ bundle that ships to every team.

Synced tables

• Underscore aliases (e.g. postgres_posthog_dashboarditem) point at the same synced data.

• These are replicas of the Django tables in this repo (posthog_dashboarditem backs the Insight model), so rows span every team; team_id is the scoping column.

• More prod tables than these are synced. Before concluding cross-team data is inaccessible, check the catalog:

  SELECT table_name, description
  FROM system.information_schema.tables
  WHERE table_type = 'data_warehouse' AND table_name ILIKE '%postgres%'
  

Workflow

1. Confirm columns before projecting — synced schemas drift with the Django models:

   SELECT column_name, data_type
   FROM system.information_schema.columns
   WHERE table_name = 'postgres.posthog_dashboarditem'
   

2. Query with posthog:execute-sql, filtering or grouping by team_id. Example — most active teams by insights created in the last 30 days: