Review Stage3A scenario outputs for realism, independence, and hazard-event logic.
npx skills add https://github.com/WSJ40/hara-mutiagent --skill hara-stage3ar
Please help me install this Agent Skill. Skill: hara-stage3ar Source: https://github.com/WSJ40/hara-mutiagent/tree/main/skills/hara-stage3ar Command: npx skills add https://github.com/WSJ40/hara-mutiagent --skill hara-stage3ar
This Skill prevents unreliable or inconsistent Stage3A scenario outputs by performing a dedicated Stage3AR review that focuses on scenario realism, independence, domain consistency, hazard-event logic, and scenario reasoning—without re-scoring S/E/C/ASIL/FTTI or overall safety state.
Ask the assistant to run Stage3A checks, prepare Stage3AR batch context from your Stage3A scenarios, then generate and merge Stage3AR review outputs for output/<RUN_ID>stage3a<MF_ID>_scenarios.json.
High-intent search queries and answers about installing and using this skill.
Stage3A scenario validation requires running check_stage_json.py --stage stage3a to verify scenario realism, independence, and hazard logic, producing a review JSON file for batch traceability.
Stage3AR scenario review checks operational domain consistency, hazard-event logic, scenario independence, and reasoning coherence between motion direction, road conditions, risk objects, and driver presence.
Generate a Stage3A review JSON by evaluating scenario batches using the stage3a-review-batches context, then merging results into output/<RUN_ID>_stage3a_<MF_ID>_review.json for traceability.
Stage3AR review explicitly avoids S/E/C/ASIL/FTTI and safety-state scoring, focusing solely on scenario realism, independence, and hazard-event logic validation between Stage3A and Stage3B.
Gatekeeping Stage3A scenarios involves running Stage3AR batch evaluations to approve corrected outputs, ensuring scenario reasoning and hazard logic consistency before allowing Stage3B rating to begin.
Stage3AR batch validation checks coverage of max_asil_planning, while also verifying the logical consistency of hazard events and internal coherence of operational domain constraints.
WSJ40Generate Stage 3B SEC rating outputs by merging S/E/C/FTTI batch files into a contract-compliant sec.json.
WSJ40Validate Stage 3B SEC ratings, FTTI consistency, and safety goal coverage.
WSJ40Orchestrate Stage0–Stage4 HARA workflows with isolated sub-agents and validation gates.
WSJ40Generate Stage 3A operational scenarios and hazard events for a single MF.
