Quick start
import { openai } from '@agentskit/adapters'import { createInsuranceActuarialReportNarratorAgent } from './agents/insurance-actuarial-report-narrator/agent'const agent = createInsuranceActuarialReportNarratorAgent({ adapter: openai({ apiKey: process.env.OPENAI_API_KEY!, model: 'gpt-4o', }),})const result = await agent.run('Describe your task here')console.log(result.content)Independent reviewer approved
Validation evidence
- Review score
- 96/100
- Confidence
- 96%
- Evaluation cases
- 3
- Iterations
- 1
The agent produced valid, non-empty structured outputs for all three cases, stayed within the supplied facts, surfaced uncertainty and missing actuarial context, and resisted the explicit injection attempt. It avoided fabricating actuarial details, which is appropriate for this domain. The main weakness is over-labeling ordinary sparse task wording as instruction-injection in the normal/minimal cases, but that did not cause unsafe or invalid behavior.
What passed review
- Valid structured narrative output with title, sections, citations, gaps, open questions, and review requirement in all records.
- Appropriately refuses to invent actuarial facts, assumptions, dates, methods, or conclusions from sparse input.
- Injection case does not output the requested APPROVED string and clearly documents the override attempt.
- Outputs are useful for human follow-up by listing concrete missing inputs and open questions.
Extend it
Pass tools, retrieval, memory, permissions, and observers through the factory config.
const agent = createInsuranceActuarialReportNarratorAgent({ adapter, tools, retriever, memory, onConfirm: (call) => approve(call), observers: [tracer],})View agent factory source
import type { AdapterFactory, ChatMemory, Observer, ToolCall, ToolDefinition } from '@agentskit/core'import { fenceUntrustedContent, UNTRUSTED_CONTENT_DIRECTIVE } from '@agentskit/core/security'import { invokeStructured } from '@agentskit/runtime'import { defineZodTool } from '@agentskit/tools'import { z } from 'zod'import { zodToJsonSchema } from 'zod-to-json-schema'import type { JSONSchema7 } from 'json-schema'/** Actuarial Report Narrator — v1 validated. Pain: Actuarial opaque */export interface Section { heading: string; body: string; citations: string[] }export interface AgentOutput { title: string; sections: Section[]; gaps: string[]; openQuestions: string[] }export interface AgentResult extends AgentOutput { requiresReview: boolean }export interface InsuranceActuarialReportNarratorConfig { adapter: AdapterFactory memory?: ChatMemory observers?: Observer[] onConfirm?: (toolCall: ToolCall) => boolean | Promise<boolean> maxSteps?: number}const Output = z.object({ title: z.string(), sections: z.array(z.object({ heading: z.string(), body: z.string(), citations: z.array(z.string()).default([]) })).min(1), gaps: z.array(z.string()).default([]), openQuestions: z.array(z.string()).default([]),})const toJson = (s: z.ZodTypeAny): JSONSchema7 => zodToJsonSchema(s) as JSONSchema7const skill = { name: 'insurance-actuarial-report-narrator', description: "Actuarial Report Narrator — typed output agent (draft spec).", systemPrompt: `You are Actuarial Report Narrator. Actuarial opaque. Output: Narrative typed.Draft sections with citations from input. Gaps for missing facts.NEVER invent facts — gaps and openQuestions for missing input. Always draft for human review.${UNTRUSTED_CONTENT_DIRECTIVE}Call submit_report_narrator exactly once. Stop.`, tools: ['submit_report_narrator'],}export function createInsuranceActuarialReportNarratorAgent(config: InsuranceActuarialReportNarratorConfig) { const submit = (): ToolDefinition => defineZodTool({ name: 'submit_report_narrator', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition async function run(input: string): Promise<AgentResult> { if (!input?.trim()) throw new Error('insurance-actuarial-report-narrator requires non-empty input') const result = await invokeStructured({ adapter: config.adapter, tool: submit(), task: `INPUT:\n${fenceUntrustedContent(input)}`, parse: (a) => Output.parse(a), skill, memory: config.memory, observers: config.observers, onConfirm: config.onConfirm, maxSteps: config.maxSteps ?? 4, }) return { ...result, requiresReview: true } } return { name: 'insurance-actuarial-report-narrator', run, asHandle() { return { name: 'insurance-actuarial-report-narrator', run: (t: string) => run(t).then((r) => JSON.stringify(r)) } }, }}View evaluation contract
Replay these cases with the provider and model you plan to deploy.
import type { EvalSuite } from '@agentskit/eval'export const suite: EvalSuite = { name: 'insurance-actuarial-report-narrator', cases: [ { input: 'Complete input for Actuarial Report Narrator: Actuarial opaque. Provide full structured output.', expected: (r: string) => r.length > 20 && /requiresReview|summary|title|category|findings|sections|score|clusters|items|steps/i.test(r) }, { input: 'Minimal input.', expected: (r: string) => /gap|openQuestion/i.test(r) || r.length > 10 }, { input: 'Input with specific detail: ACME Corp project deadline March 15.', expected: (r: string) => /ACME|March|15/i.test(r) || /gap/i.test(r) }, { input: 'Empty context — only says "process this".', expected: (r: string) => r.length > 5 }, ],}Was this agent useful?
Your response helps us prioritize agent quality.