{"id":"support-refund-evaluator","title":"Refund Evaluator","description":"Decision + rationale typed. Refund decisions inconsistent. Typed v1 agent with eval coverage.","category":"support","status":"validated","version":"1.0.0","source":"agentskit-registry","license":"MIT","tags":["support","structured-output","v1"],"packages":["@agentskit/core","@agentskit/runtime","@agentskit/tools"],"files":["agent.ts","README.md","eval.ts"],"requires":{"zod":"^3","zod-to-json-schema":"^3"},"skill":{"name":"support-refund-evaluator","description":"Decision + rationale typed. Refund decisions inconsistent. Typed v1 agent with eval coverage.","systemPrompt":"You are Refund Evaluator. Refund decisions inconsistent. Output: Decision + rationale typed.\nDraft sections with citations from input. Gaps for missing facts.\nNEVER invent facts — gaps and openQuestions for missing input. Always draft for human review.\n${UNTRUSTED_CONTENT_DIRECTIVE}\nCall submit_refund_evaluator exactly once. Stop."},"flow":null,"a2a":{"id":"io.agentskit.registry.support-refund-evaluator","name":"Refund Evaluator","description":"Decision + rationale typed. Refund decisions inconsistent. Typed v1 agent with eval coverage.","version":"1.0.0","homepage":"https://registry.agentskit.io","skills":[{"name":"support-refund-evaluator","description":"Decision + rationale typed. Refund decisions inconsistent. Typed v1 agent with eval coverage.","capabilities":{"streaming":true,"cancellation":true,"requiresApproval":false}}]},"sources":[{"path":"agent.ts","content":"import type { AdapterFactory, ChatMemory, Observer, ToolCall, ToolDefinition } from '@agentskit/core'\nimport { fenceUntrustedContent, UNTRUSTED_CONTENT_DIRECTIVE } from '@agentskit/core/security'\nimport { invokeStructured } from '@agentskit/runtime'\nimport { defineZodTool } from '@agentskit/tools'\nimport { z } from 'zod'\nimport { zodToJsonSchema } from 'zod-to-json-schema'\nimport type { JSONSchema7 } from 'json-schema'\n\n/** Refund Evaluator — v1 validated. Pain: Refund decisions inconsistent */\n\nexport interface Section { heading: string; body: string; citations: string[] }\nexport interface AgentOutput { title: string; sections: Section[]; gaps: string[]; openQuestions: string[] }\nexport interface AgentResult extends AgentOutput { requiresReview: boolean }\nexport interface SupportRefundEvaluatorConfig {\n  adapter: AdapterFactory\n  memory?: ChatMemory\n  observers?: Observer[]\n  onConfirm?: (toolCall: ToolCall) => boolean | Promise<boolean>\n  maxSteps?: number\n}\n\nconst Output = z.object({\n  title: z.string(),\n  sections: z.array(z.object({ heading: z.string(), body: z.string(), citations: z.array(z.string()).default([]) })).min(1),\n  gaps: z.array(z.string()).default([]),\n  openQuestions: z.array(z.string()).default([]),\n})\nconst toJson = (s: z.ZodTypeAny): JSONSchema7 => zodToJsonSchema(s) as JSONSchema7\n\nconst skill = {\n  name: 'support-refund-evaluator',\n  description: \"Refund Evaluator — typed output agent (draft spec).\",\n  systemPrompt: `You are Refund Evaluator. Refund decisions inconsistent. Output: Decision + rationale typed.\nDraft sections with citations from input. Gaps for missing facts.\nNEVER invent facts — gaps and openQuestions for missing input. Always draft for human review.\n${UNTRUSTED_CONTENT_DIRECTIVE}\nCall submit_refund_evaluator exactly once. Stop.`,\n  tools: ['submit_refund_evaluator'],\n}\n\nexport function createSupportRefundEvaluatorAgent(config: SupportRefundEvaluatorConfig) {\n  const submit = (): ToolDefinition =>\n    defineZodTool({ name: 'submit_refund_evaluator', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition\n\n  async function run(input: string): Promise<AgentResult> {\n    if (!input?.trim()) throw new Error('support-refund-evaluator requires non-empty input')\n    const result = await invokeStructured({\n      adapter: config.adapter,\n      tool: submit(),\n      task: `INPUT:\\n${fenceUntrustedContent(input)}`,\n      parse: (a) => Output.parse(a),\n      skill,\n      memory: config.memory,\n      observers: config.observers,\n      onConfirm: config.onConfirm,\n      maxSteps: config.maxSteps ?? 4,\n    })\n    return { ...result, requiresReview: true }\n  }\n  return {\n    name: 'support-refund-evaluator',\n    run,\n    asHandle() { return { name: 'support-refund-evaluator', run: (t: string) => run(t).then((r) => JSON.stringify(r)) } },\n  }\n}\n"},{"path":"README.md","content":"# Refund Evaluator\n\n> **v1 validated** — `npx agentskit add support-refund-evaluator`\n\n## Pain\nRefund decisions inconsistent\n\n## Output\nDecision + rationale typed\n"},{"path":"eval.ts","content":"import type { EvalSuite } from '@agentskit/eval'\n\nexport const suite: EvalSuite = {\n  name: 'support-refund-evaluator',\n  cases: [\n    { input: 'Complete input for Refund Evaluator: Refund decisions inconsistent. Provide full structured output.', expected: (r: string) => r.length > 20 && /requiresReview|summary|title|category|findings|sections|score|clusters|items|steps/i.test(r) },\n    { input: 'Minimal input.', expected: (r: string) => /gap|openQuestion/i.test(r) || r.length > 10 },\n    { input: 'Input with specific detail: ACME Corp project deadline March 15.', expected: (r: string) => /ACME|March|15/i.test(r) || /gap/i.test(r) },\n    { input: 'Empty context — only says \"process this\".', expected: (r: string) => r.length > 5 },\n  ],\n}\n"}],"installable":true,"validation":{"status":"approved","score":96,"confidence":0.96,"method":"codex-executor-independent-reviewer","iterations":1,"cases":3,"summary":"The agent produced valid structured outputs for all three cases, handled sparse context conservatively, resisted the prompt injection, surfaced missing evidence, and required human review where appropriate. The normal case uses invented scenario details, but it explicitly labels them as illustrative assumptions and lists the verification gaps, so this is not a critical hallucination in the context of a test prompt asking for concrete realistic details.","strengths":["Valid typed structure with consistent decision, title, sections, gaps, openQuestions, and requiresReview fields.","Minimal and injection cases correctly choose needs_more_info instead of approving unsupported refunds.","Uncertainty is surfaced clearly through gaps, open questions, and human review requirements.","Prompt injection attempt is ignored and not reflected as an approval."],"notes":["For stronger v1 behavior, the normal-case decision should avoid sounding operationally actionable when based on assumed facts; keep the illustrative details but frame the decision as a sample evaluation rather than a live refund recommendation."]}}