hara-orchestrator

End-to-end HARA generation is automated by dispatching isolated per-Function and per-MF sub-agent work across Stage0 through Stage4, running deterministic validation checks between stages to ensure safety analysis consistency before final export.

Yes, specific failed stages in the multi-agent safety analysis workflow can be rerun individually because the orchestrator runs deterministic validation gates between stages and supports recovery without executing the entire pipeline from scratch.

Multi-agent orchestration workflows enforce JSON-only communication by ensuring each stage and review process runs in properly isolated sub-agents that exchange artifacts exclusively through validated JSON schemas, satisfying strict inter-agent communication requirements.

Yes, the HARA orchestration workflow exports final deliverables to both JSON and Excel formats, generating structured output files in your designated output directory after Stage4 SG_Sum derivation and SEC evaluation are complete.

The best way to handle Stage3A and Stage3B scenario generation is using a gated orchestration pipeline that dispatches isolated sub-agent work for SEC evaluation, followed by Stage4 SG_Sum derivation with stop-on-validation-error behavior.

Multi-agent workflow automation requires stage-specific schema routing to ensure deterministic validation gates can accurately verify JSON artifacts between Stage0 and Stage4, preventing invalid safety analysis data from propagating downstream.