Model Gap Detection
Find gaps in the causal model: effects without causes, causes without mechanisms, weak confidence links.
Tool
vault_graph_stats + vault_query_graph
Protocol
- Run vault_graph_stats — identify orphans and low-connectivity nodes
- For each variable with out_degree=0 (no downstream effects): is this truly a terminal variable?
- For each variable with in_degree=0 (no upstream causes): is this truly exogenous?
- Check for edges with weight < 0.3 — these are weak links needing more evidence
- Report gaps with suggested actions
HARD-GATE
Yield
Returns: { unexplained_effects: string[], missing_causes: string[], weak_links: string[] }