{"id":"bench-orch-001","slug":"bench-orch-001","title":"Route research then code fix","prompt":"Orchestrate: (1) research agent finds the correct algorithm name for problem brief, (2) coding agent implements it in stub file, (3) merge into final answer. Log assignments.","successCriteria":["Final tests pass","Assignment log shows ≥2 specialists","No duplicate full reimplementation by orchestrator alone if specialists available"],"scoringHints":["Collaboration + planning"],"agentTypes":["multi_agent_orchestrator"],"primaryType":"multi_agent_orchestrator","dimensions":["collaborationHandoff","planningDecomposition","taskSuccess","efficiency"],"difficulty":"medium","timeBudgetSec":1200,"tags":["routing","merge"],"needsTools":true,"needsWeb":false,"needsCode":true,"holdout":false,"adversarialFlags":[],"type":{"id":"multi_agent_orchestrator","slug":"multi_agent_orchestrator","name":"Multi-agent orchestrator","definition":"Controller that decomposes work, routes to specialist agents/tools, merges results, resolves conflicts, and produces a coherent final artifact with clear ownership.","typicalTools":["agent registry","message bus / A2A","shared memory","router/planner","critic/verifier agents","budget governor"],"whatGoodLooksLike":"Correct merge without duplicated work; explicit handoffs; conflict detection; budget adherence; failure isolation; auditable assignment log.","dimensionWeightsHigh":["collaborationHandoff","planningDecomposition","taskSuccess","efficiency","multiStepRobustness","honestyCalibration"],"relatedBenchmarks":["Multi-agent coordination evals; ARE/Gaia2 multi-actor vibes"],"exampleJtbds":["durable-agent-workflows","langgraph-deploy","agent-tool-router","stateful-agent-runtime"],"dimensionWeights":{"taskSuccess":0.15,"toolUseQuality":0.06,"planningDecomposition":0.14,"reliabilityConsistency":0.05,"honestyCalibration":0.06,"safetyPolicy":0.05,"efficiency":0.08,"memoryContext":0.05,"instructionFollowing":0.05,"multiStepRobustness":0.08,"collaborationHandoff":0.15,"groundingEvidence":0.03,"communicationClarity":0.03,"adaptivity":0.02},"defaultTaskIds":["bench-orch-001","bench-orch-002","bench-orch-003","bench-orch-004","bench-orch-005","bench-cross-005","bench-orch-006","bench-orch-007","bench-orch-008"]},"gold":null,"note":"Holdout gold is never returned. Grade via traces + verifiers."}