validation-patterns

Call workflow.build() before runtime.execute(), which raises WorkflowValidationError for invalid parameters or connections. You can also use the WorkflowValidator class with validate_all() to collect errors and warnings in one pass.

Connections must use the 4-parameter format: add_connection(source_id, source_param, target_id, target_param). Validation checks that both nodes exist, parameter names match node inputs and outputs, and no circular dependencies exist.

Check for hardcoded secrets, SQL injection via string concatenation in queries, code injection in PythonCodeNode, and file path traversal. Use grep patterns to find hardcoded API keys and prefer environment variables and parameterized queries.

Common causes include missing required node parameters, wrong parameter types, unknown parameters not in the node contract, missing .build() before execute, or 3-parameter connections instead of the required 4-parameter format.

Yes, DataFlow results must be accessed as results["node_id"]["result"], not attribute access like results["node_id"].result. Validation also checks string ID preservation, one DataFlow instance per database, and proper @db.model decorator usage.