{
  "id": "workflow:agent-evaluation-cycle",
  "_kind": "Workflow",
  "_file": "domain/workflows/workflows-technical-depth.yaml",
  "_cluster": "domain",
  "attributes": {
    "displayName": "Agent Evaluation Cycle",
    "description": "Rigorous evaluation workflow for measuring the accuracy, reliability, and safety of\nAI agent systems across defined benchmark tasks and adversarial scenarios. The ML\nengineer assembles an evaluation harness with a curated dataset of prompts, expected\noutputs, and rubric-based scoring functions. The backend engineer integrates the\nharness into CI so every model or prompt change triggers an automated eval run.\nRegression thresholds enforce that new versions do not degrade on prior benchmarks,\nwhile exploratory eval sessions probe edge cases and failure modes that inform the\nnext iteration of the agent's architecture or system prompt.\n",
    "workflowKind": "development",
    "triggerType": "on-demand",
    "typicalCadence": "per-sprint",
    "complexity": "complex"
  },
  "outgoingEdges": [
    {
      "from": "workflow:agent-evaluation-cycle",
      "to": "role:ml-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:agent-evaluation-cycle",
      "to": "role:backend-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:agent-evaluation-cycle",
      "to": "role:research-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:agent-evaluation-cycle",
      "to": "role:qa-engineer",
      "kind": "involves_role"
    },
    {
      "from": "workflow:agent-evaluation-cycle",
      "to": "domain:software-engineering",
      "kind": "applies_to_domain"
    }
  ],
  "incomingEdges": [
    {
      "from": "stack-profile:multi-agent-orchestration",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "follows_workflow"
    },
    {
      "from": "stack-profile:voice-ai-agent",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "follows_workflow"
    },
    {
      "from": "stack-profile:autonomous-agent-fleet",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "follows_workflow"
    },
    {
      "from": "stack-profile:prompt-engineering-workbench",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "follows_workflow"
    },
    {
      "from": "tool:fireworks-ai",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Model API outputs can be compared across prompts, eval suites, and agent runs."
      }
    },
    {
      "from": "tool:mistral",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Model outputs can be benchmarked during agent evaluation cycles."
      }
    },
    {
      "from": "tool:openai",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "high",
        "evidence": "Model responses and tool-use traces can be measured through agent evaluation cycles."
      }
    },
    {
      "from": "tool:deepseek",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Agent and coding-benchmark outputs can be compared against other LLM providers."
      }
    },
    {
      "from": "tool-server:mcp-mistral-ai-candidate",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Model outputs can be benchmarked during agent evaluation."
      }
    },
    {
      "from": "tool-server:mcp-deepseek-candidate",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "supports_work",
      "attributes": {
        "confidence": "medium",
        "evidence": "Coding-agent outputs can be measured in evaluation cycles."
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--ab-testing-conversational",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 0.7
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--add-app-to-mcp-server",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--advanced-rag-patterns",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--agent-evaluation-framework",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--agent-evaluation-framework",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 0.7
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--agent-performance-optimization",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--autonomous-task-planning",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--bias-detection-fairness",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--content-moderation-safety",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--conversational-memory-system",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--convert-web-app-to-mcp",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--create-mcp-app",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--custom-tool-development",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--empathetic-response-generation",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--entity-extraction-slot-filling",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--intent-classification-system",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--knowledge-base-qa",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--llm-fine-tuning-conversational",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--llm-observability-monitoring",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--long-term-memory-management",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--multi-agent-system",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--multi-modal-agent",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--prompt-engineering-workflow",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--prompt-injection-defense",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--react-agent-implementation",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--regression-testing-agent",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 0.7
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--self-reflection-agent",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--system-prompt-guardrails",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--tool-safety-validation",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    },
    {
      "from": "lib-process:ai-agents-conversational--voice-enabled-conversational",
      "to": "workflow:agent-evaluation-cycle",
      "kind": "lib_implements_workflow",
      "attributes": {
        "weight": 1
      }
    }
  ]
}