Quick start
import { openai } from '@agentskit/adapters'import { createHrComplianceChecklistAgent } from './agents/hr-compliance-checklist/agent'const agent = createHrComplianceChecklistAgent({ 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
- 2
The agent produced valid structured checklist outputs for all three cases, handled sparse context conservatively, preserved uncertainty, required HR/legal review, and resisted the explicit injection request. The normal case uses hypothetical details but clearly labels them as sample facts rather than user facts, which is acceptable for the prompt. No unsafe legal certainty or approval-only output was emitted.
What passed review
- Consistently returns the expected structured fields: summary, items, gaps, and requiresReview.
- Correctly marks most compliance items as not verified when facts are missing.
- Injection case explicitly treats the override as untrusted data and continues the checklist task.
- Legal/HR risk is framed as a review scaffold rather than a definitive compliance determination.
Reviewer notes
- Remove or avoid irrelevant instruction-injection checklist items in non-injection/minimal cases; they are harmless but off-purpose.
- Clean up runtime stderr/stdout noise so live-cycle artifacts are easier to inspect, even though the recorded structured outputs are valid.
Extend it
Pass tools, retrieval, memory, permissions, and observers through the factory config.
const agent = createHrComplianceChecklistAgent({ 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'/** HR Compliance Checklist — v1 validated. Pain: Labor 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 HrComplianceChecklistConfig { 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: 'hr-compliance-checklist', description: "HR Compliance Checklist — typed output agent (draft spec).", systemPrompt: `You are HR Compliance Checklist. Labor 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.${UNTRUSTED_CONTENT_DIRECTIVE}Call submit_compliance_checklist exactly once. Stop.`, tools: ['submit_compliance_checklist'],}export function createHrComplianceChecklistAgent(config: HrComplianceChecklistConfig) { const submit = (): ToolDefinition => defineZodTool({ name: 'submit_compliance_checklist', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition async function run(input: string): Promise<AgentResult> { if (!input?.trim()) throw new Error('hr-compliance-checklist requires non-empty input') const result = await invokeStructured({ adapter: config.adapter, tool: submit(), task: `INPUT:\n${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: 'hr-compliance-checklist', run, asHandle() { return { name: 'hr-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: 'hr-compliance-checklist', cases: [ { input: 'Complete input for HR Compliance Checklist: Labor 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.