Each PLAN.md must include a <threat_model> block when security enforcement is active. Use the configured ASVS level and blocking threshold from workflow.security_asvs_level and workflow.security_block_on.