{"product_id":"anthropic-claude-sonnet-agent-reasoning-quality-scoring-n8n","title":"Anthropic Claude Sonnet Agent Reasoning Quality Scoring (n8n)","description":"\u003ch3\u003eScore an agent’s reasoning trace with Claude Sonnet—then pass\/fail it automatically\u003c\/h3\u003e\n\u003cp\u003eThis \u003cstrong\u003en8n automation workflow\u003c\/strong\u003e scores the \u003cem\u003equality of an agent’s reasoning trace\u003c\/em\u003e using \u003cstrong\u003eAnthropic Claude Sonnet 4.6\u003c\/strong\u003e. It generates per-dimension quality scores with justifications and returns a \u003cstrong\u003epassed \/ failed\u003c\/strong\u003e result based on configurable thresholds—so your automation can reliably decide whether an agent’s output is acceptable.\u003c\/p\u003e\n\n\u003ch3\u003eWhat this workflow does\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003e\n\u003cstrong\u003eReceives inputs\u003c\/strong\u003e from a parent workflow call: \u003ccode\u003ereasoning_trace\u003c\/code\u003e (array), the agent \u003ccode\u003everdict\u003c\/code\u003e, and \u003ccode\u003eraw_evidence\u003c\/code\u003e.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eApplies quality thresholds\u003c\/strong\u003e by setting an \u003cstrong\u003eoverall reasoning quality threshold\u003c\/strong\u003e and a \u003cstrong\u003eper-dimension threshold\u003c\/strong\u003e to define pass\/fail criteria.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eCalls Anthropic Claude Sonnet 4.6\u003c\/strong\u003e to score the trace and evidence across dimensions such as:\n    \u003cul\u003e\n      \u003cli\u003e\u003cstrong\u003eEvidence grounding\u003c\/strong\u003e\u003c\/li\u003e\n      \u003cli\u003e\u003cstrong\u003eCircularity\u003c\/strong\u003e\u003c\/li\u003e\n      \u003cli\u003e\u003cstrong\u003eContradiction handling\u003c\/strong\u003e\u003c\/li\u003e\n      \u003cli\u003e\u003cstrong\u003eConfidence calibration\u003c\/strong\u003e\u003c\/li\u003e\n    \u003c\/ul\u003e\n  \u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eExtracts structured results\u003c\/strong\u003e (dimension scores, justifications, and the overall \u003ccode\u003ereasoning_quality\u003c\/code\u003e value) from Claude’s JSON output.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eEvaluates thresholds\u003c\/strong\u003e and returns:\n    \u003cul\u003e\n      \u003cli\u003e\u003ccode\u003epassed: true\/false\u003c\/code\u003e\u003c\/li\u003e\n      \u003cli\u003e\n\u003ccode\u003eflagged_dimensions\u003c\/code\u003e listing any dimension scores at or below threshold\u003c\/li\u003e\n    \u003c\/ul\u003e\n  \u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eReturns a merged result\u003c\/strong\u003e back to the calling workflow for downstream automation.\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003ch3\u003eUse cases\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003e\n\u003cstrong\u003eQuality gating\u003c\/strong\u003e for AI agents in n8n: reject weak or poorly grounded reasoning before acting.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003eAudit-friendly evaluation\u003c\/strong\u003e: capture justifications per dimension for review and iteration.\u003c\/li\u003e\n  \u003cli\u003e\n\u003cstrong\u003ePolicy enforcement\u003c\/strong\u003e: automatically fail outputs when contradictions or circular logic are detected.\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003ch3\u003eTechnical details\u003c\/h3\u003e\n\u003cul\u003e\n  \u003cli\u003eRuns as a \u003cstrong\u003esub-workflow\u003c\/strong\u003e called by an \u003cstrong\u003eExecute Workflow\u003c\/strong\u003e step in a parent workflow.\u003c\/li\u003e\n  \u003cli\u003eUses nodes including \u003ccode\u003eif\u003c\/code\u003e, \u003ccode\u003eset\u003c\/code\u003e, and \u003ccode\u003emerge\u003c\/code\u003e, along with an \u003ccode\u003eexecute workflow trigger\u003c\/code\u003e structure.\u003c\/li\u003e\n  \u003cli\u003eUses the \u003cstrong\u003eAnthropic Chat Model\u003c\/strong\u003e via an \u003cstrong\u003eAnthropic credential\u003c\/strong\u003e for \u003cstrong\u003eClaude Sonnet 4.6\u003c\/strong\u003e.\u003c\/li\u003e\n  \u003cli\u003eIncludes UI elements such as a \u003cstrong\u003esticky note\u003c\/strong\u003e for setup guidance and configuration.\u003c\/li\u003e\n\u003c\/ul\u003e\n\n\u003cp\u003e\u003cstrong\u003eSetup tip:\u003c\/strong\u003e Create\/select an Anthropic credential, then configure \u003ccode\u003eoverall_threshold\u003c\/code\u003e and \u003ccode\u003edimension_threshold\u003c\/code\u003e in the configuration step to match your quality bar.\u003c\/p\u003e","brand":"N8N Commerce","offers":[{"title":"Default Title","offer_id":45895373324467,"sku":"N8N-18651","price":52.99,"currency_code":"GBP","in_stock":true}],"thumbnail_url":"\/\/cdn.shopify.com\/s\/files\/1\/0749\/6279\/6723\/files\/CQEEt0pXxDpMHpYyBsjMT_1rzFRbbW.png?v=1787649149","url":"https:\/\/buyflowscripts.com\/products\/anthropic-claude-sonnet-agent-reasoning-quality-scoring-n8n","provider":"N8N Commerce","version":"1.0","type":"link"}