Schema Analysis

Schema analysis is the first step after connecting your Supabase project. Skene reads your database structure, identifies key patterns, and builds a foundation for feature compilation and trigger deployment.

What it does

Skene introspects your Postgres information_schema to understand:

• Tables and columns — names, types, constraints, primary keys
• Foreign key relationships — how tables reference each other
• Schema boundaries — which schemas are available (public, auth, custom schemas)

From this raw structure, Skene builds a TTV (Time-Trigger-Value) map and suggests lifecycle stages for your product.

Running analysis

From the dashboard

Workspace home (/workspace/<slug>) is the main overview: run analysis, explore the TTV (Time-Trigger-Value) journey graph, filter by subject, and use the overview dock. The flow is typically:

1. Precheck — Verify Supabase connection and schema access
2. Enrich — Optionally pull context from a linked GitHub repository to improve table labeling
3. Seed — Generate lifecycle stage suggestions from the schema
4. Compile — Build the TTV map and prepare for feature creation

The Agent Engine page (/workspace/<slug>/skene-engine) offers the same graph plus deeper engine tools: compiled loops, engine YAML, GitHub sync, and push artifacts — use it when you are iterating on the engine and manifests alongside schema analysis.

From the CLI

# Analyze your codebase (produces growth-manifest.json)
uvx skene analyze .

# Push results to Skene Cloud
uvx skene push

The CLI analyzes your codebase (not the database directly) and pushes the manifest to Cloud. Cloud can then cross-reference CLI analysis with its own schema introspection.

TTV map

The TTV (Time-Trigger-Value) map scores each table across three dimensions:

Tables with high TTV scores are strong candidates for growth triggers — they represent meaningful user actions and state changes.

Lifecycle stages

Based on the TTV analysis, Skene suggests lifecycle stages for your product. These represent the journey a user takes:

Trial → Onboarded → Active → At Risk → Churned

How stages are assigned

Skene maps table states to lifecycle stages using field names and patterns:

• A status = 'trial' field maps to the Trial stage
• An onboarding_completed_at timestamp maps to Onboarded
• Recent activity timestamps map to Active
• Absence of recent activity maps to At Risk

Overriding stages

On the journey graph (workspace home or Agent Engine), you can:

• Reassign tables to different lifecycle stages by clicking nodes in the graph
• Add or remove lifecycle stages from the stage panel
• Reorder stages to match your product's actual user journey

Primary subject detection

Skene identifies the primary subject — the entity whose lifecycle you're tracking. This is typically a user, account, or organization.

Detection uses:

• Foreign key relationships (which table do others reference?)
• Column naming patterns (user_id, account_id, org_id)
• Table structure (the auth.users table in Supabase projects)

The primary subject determines how features resolve their target entity when a trigger fires.

How analysis feeds into features

Schema analysis produces the context that feature compilation needs:

1. Table mappings — Which table and operation a feature should trigger on
2. Column selection — Which columns to include in trigger payloads (sensitive columns like passwords and tokens are automatically excluded)
3. Entity resolution — How to find the user or entity ID from a triggered row
4. Enrichment — Which related tables to join for additional context

When you create a feature, Skene uses this analysis to automatically suggest the best table, columns, and entity mapping. You can override any suggestion.

Next steps

• Features & Deploy — Create features using the schema analysis results
• Supabase Integration — Schema setup and trigger deployment details