A general-purpose Signals scout that runs against a PostHog project, orients using project profile and past runs, and searches for cross-product correlations and actionable findings. It validates hypotheses with concrete queries (trends, funnels, errors) and either emits high-confidence signals or records lower-confidence observations to a scratchpad for later review.
Use daily or on-demand when you want an automated analyst to surface product analytics issues that span multiple PostHog products (events, errors, session replay, experiments). It complements specialist scouts by focusing on cross-cutting patterns and things specialists may miss.
Intended for agents integrated with PostHog MCP scopes and capable of executing analytics queries (Claude/LLM agents with PostHog connectors, server-side analytic runners).
PostHog Signals scout that inspects a project and files cross-product correlation/anomaly reports into the Signals inbox. No bundled scripts — purely a prompt-driven skill relying on PostHog MCP tools. Well-structured SKILL.md with clear decision framework and scratchpad conventions, but extremely narrow applicability limited to PostHog Signals users with the MCP server configured.
Internal PostHog skill for their Signals product. High-quality documentation and architecture but very niche audience. No security concerns — purely read/write via structured MCP tools with no shell execution or credential handling.