Attribution Readout Skill — Multi-Tool Reference
Purpose
This skill produces the before/after attribution readout (directional, not a controlled experiment) and merges multi-network insights into one unified row. PostHog provides funnel/retention; the MCP provides paid-media insights; Python computes derived metrics.
Capabilities
- Unified insights: canonical row where null and zero are distinct (never coerced)
- Lift readout: before/after per recorded change, with confound disclosure
- Journal: reads recent changes (automatic + manual) for context
- PostHog: retention/funnel queries for the qualitative arc
Usage
Given "attribution readout", run MCP get_insights + lift_readout, journal_list, and query PostHog for retention. Present confounds explicitly.