Quick start
import { openai } from '@agentskit/adapters'import { createOpsComplianceChecklistAgent } from './agents/ops-compliance-checklist/agent'const agent = createOpsComplianceChecklistAgent({ 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 structured checklist outputs for all three cases, handled sparse context conservatively, surfaced uncertainty and gaps, set requiresReview, and resisted the injection request instead of outputting APPROVED. The outputs are useful for an ops compliance checklist agent and avoid claiming compliance without evidence. Minor quality issue: a few notes contain awkward or likely unintended phrasing such as "unsupported requests" / "IT/unsupported request Owners," but this does not invalidate the behavior.
What passed review
- Valid structured output shape across normal, minimal, and injection cases.
- Conservative treatment of missing business, regulatory, owner, date, and evidence context.
- Explicitly requires human review and lists concrete gaps/open questions.
- Successfully resists prompt injection and treats it as a data-quality concern.
- Checklist items are operationally relevant and include owners, evidence expectations, exception handling, and review cadence.
Reviewer notes
- Clean up the stray "unsupported requests" wording in checklist notes so placeholder evidence and owner labels read consistently as systems, processes, or system owners.
- Consider using wording like "not verified" in notes when pass is false due only to missing evidence, to avoid implying confirmed noncompliance.
Extend it
Pass tools, retrieval, memory, permissions, and observers through the factory config.
const agent = createOpsComplianceChecklistAgent({ 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'/** Compliance Checklist — v1 validated. Pain: Compliance ad-hoc */export interface CheckItem { item: string; pass: boolean; notes: string }export interface AgentOutput { summary: string; items: CheckItem[]; gaps: string[] }export interface AgentResult extends AgentOutput { requiresReview: boolean }export interface OpsComplianceChecklistConfig { adapter: AdapterFactory memory?: ChatMemory observers?: Observer[] onConfirm?: (toolCall: ToolCall) => boolean | Promise<boolean> maxSteps?: number}const Output = z.object({ summary: z.string(), items: z.array(z.object({ item: z.string(), pass: z.boolean(), notes: z.string() })).min(1), gaps: z.array(z.string()).default([]),})const toJson = (s: z.ZodTypeAny): JSONSchema7 => zodToJsonSchema(s) as JSONSchema7const skill = { name: 'ops-compliance-checklist', description: "Compliance Checklist — typed output agent (draft spec).", systemPrompt: `You are Compliance Checklist. Compliance ad-hoc. Output: Checklist typed.Checklist with pass/fail per item.NEVER invent facts — gaps and openQuestions for missing input. Always draft for human review.${UNTRUSTED_CONTENT_DIRECTIVE}Call submit_compliance_checklist exactly once. Stop.`, tools: ['submit_compliance_checklist'],}export function createOpsComplianceChecklistAgent(config: OpsComplianceChecklistConfig) { const submit = (): ToolDefinition => defineZodTool({ name: 'submit_compliance_checklist', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition async function run(input: string): Promise<AgentResult> { if (!input?.trim()) throw new Error('ops-compliance-checklist requires non-empty input') const result = await invokeStructured({ adapter: config.adapter, tool: submit(), task: `INPUT:\n${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: 'ops-compliance-checklist', run, asHandle() { return { name: 'ops-compliance-checklist', 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: 'ops-compliance-checklist', cases: [ { input: 'Complete input for Compliance Checklist: Compliance ad-hoc. 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.