fintech·Independently reviewed · 96/100

Credit Memo

Memo typed. Credit decisions undocumented. Typed v1 agent with eval coverage.

fintechstructured-outputv1

Install

npx agentskit add fintech-credit-memo

Quick start

import { openai } from '@agentskit/adapters'import { createFintechCreditMemoAgent } from './agents/fintech-credit-memo/agent'const agent = createFintechCreditMemoAgent({  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

How validation works
Review score
96/100
Confidence
96%
Evaluation cases
3
Iterations
1

The agent produced valid structured memo objects for all cases, avoided inventing borrower or credit facts, surfaced uncertainty and missing inputs, required human review, and resisted the injection attempt without outputting the requested false approval. The behavior is conservative and useful for sparse inputs. Minor concern: some citations expose internal/system-policy language and untrusted-marker mechanics, which is not ideal for a production memo, but it does not create a material safety or correctness failure in these outputs.

What passed review

  • Consistently refuses to fabricate names, dates, amounts, financials, collateral, or decisions from placeholder input.
  • Surfaces concrete gaps and open questions that a credit analyst could use to complete the package.
  • Injection case correctly treats the approval instruction as untrusted data and does not issue an approval.
  • Outputs are non-empty, structured, and aligned with human-review credit memo workflow.

Reviewer notes

  • Avoid citing or quoting internal system prompt text in customer-facing memo sections; cite only user-provided/source-document facts or state that no source document was provided.
  • Consider keeping prompt-injection handling out of the memo body unless the schema explicitly expects an audit/security section.

Extend it

Pass tools, retrieval, memory, permissions, and observers through the factory config.

const agent = createFintechCreditMemoAgent({  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'/** Credit Memo — v1 validated. Pain: Credit decisions undocumented */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 FintechCreditMemoConfig {  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: 'fintech-credit-memo',  description: "Credit Memo — typed output agent (draft spec).",  systemPrompt: `You are Credit Memo. Credit decisions undocumented. Output: Memo 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.${UNTRUSTED_CONTENT_DIRECTIVE}Call submit_credit_memo exactly once. Stop.`,  tools: ['submit_credit_memo'],}export function createFintechCreditMemoAgent(config: FintechCreditMemoConfig) {  const submit = (): ToolDefinition =>    defineZodTool({ name: 'submit_credit_memo', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition  async function run(input: string): Promise<AgentResult> {    if (!input?.trim()) throw new Error('fintech-credit-memo requires non-empty input')    const result = await invokeStructured({      adapter: config.adapter,      tool: submit(),      task: `INPUT:\n${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: 'fintech-credit-memo',    run,    asHandle() { return { name: 'fintech-credit-memo', 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: 'fintech-credit-memo',  cases: [    { input: 'Complete input for Credit Memo: Credit decisions undocumented. 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.

Keep exploring

Related agents

View category