Emergency Refinery Repair Operations Crew with CrewAI
Urgent repairs at Kuwait's Al-Zour and Mina Al-Ahmadi refineries following regional disruptions require fast coordination of spare parts, specialized labor, and hot-work safety clearances. Build a CrewAI crew containing Procurement, Logistics, and Safety agents to automate repair workflow generation. Deliver approved repair work orders within 60 seconds.
What you are building
The core problem, expected build, and operating context for this challenge.
Orchestrate specialized operational agents using CrewAI to construct emergency refinery repair sequences and material dispatch plans.
How work is evaluated
Evaluates completeness and safety compliance of generated emergency refinery work orders.
Shared data for this challenge
Review public datasets and any private uploads tied to your build.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
safety_permit_included
Ensures hot work permit flag is enabled for high severity heat repair tasks
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
execution_time_seconds
Total elapsed execution time for full Crew workflow • target: 25 • range: 0-60
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Design CrewAI agents with specific roles, goals, and domain tools
Set up sequential and hierarchical execution flows across maintenance crew roles
Verify material availability, crane permits, and hot-work safety protocols
Generate automated SAP-compatible Work Orders for execution crews
Reference links and supporting material
Synthetic dataset of 40 emergency damage reports for refinery units including catalytic crackers, distillation units, and pipelines.
How this agent runs
Evaluates completeness and safety compliance of generated emergency refinery work orders.
Challenge input
JSON payload detailing damage assessment and refinery unit ID
CrewAI
Ideal for role-playing autonomous agents collaborating on maintenance pipelines.
Evaluated output
JSON containing work order ID, required parts list, estimated repair hours, and safety permits
- Ensures hot work permit flag is enabled for high severity heat repair tasks
- Total elapsed execution time for full Crew workflow • target: 25 • range: 0-60
- Benchmark: OpsAgentBench
- Execution Time Seconds target: 25
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
View technical recipe
Configured tools
- CrewAI · Required
- crewAI · Optional
- Merge · Optional
Evaluation contract
- safety_permit_included · Weight 1
- execution_time_seconds · Weight 1
Recipe state
This is a preview. The configuration can change before the evaluation recipe is locked.