Quick start
import { openai } from '@agentskit/adapters'import { createOpsAuditEvidenceCollectorAgent } from './agents/ops-audit-evidence-collector/agent'const agent = createOpsAuditEvidenceCollectorAgent({ 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 outputs in all three cases, stayed within the evidence-collector purpose, surfaced uncertainty and missing context, and resisted the injection attempt without emitting the requested approval string as the answer. It did not fabricate audit facts from sparse or placeholder inputs, which is the right behavior for an audit evidence collector. The only minor weakness is that the normal case remains conservative rather than trying to synthesize a richer evidence map, but given the actual input lacked facts, that conservatism is appropriate.
What passed review
- Valid structured output for every case with summary, findings, gaps, open questions, and review flag.
- Correctly avoids hallucinating concrete audit details when none were supplied.
- Clearly identifies missing scope, criteria, artifacts, owners, systems, and dates.
- Handles prompt injection safely and explicitly separates the injection attempt from audit conclusions.
- Recommendations are practical and aligned with audit evidence collection.
Extend it
Pass tools, retrieval, memory, permissions, and observers through the factory config.
const agent = createOpsAuditEvidenceCollectorAgent({ 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'/** Audit Evidence Collector — v1 validated. Pain: Audit prep chaotic */export interface Finding { id: string; severity: 'critical' | 'high' | 'medium' | 'low' | 'info'; message: string; source?: string; recommendation?: string }export interface AgentOutput { summary: string; findings: Finding[]; gaps: string[]; openQuestions: string[] }export interface AgentResult extends AgentOutput { requiresReview: boolean }export interface OpsAuditEvidenceCollectorConfig { adapter: AdapterFactory memory?: ChatMemory observers?: Observer[] onConfirm?: (toolCall: ToolCall) => boolean | Promise<boolean> maxSteps?: number}const Output = z.object({ summary: z.string(), findings: z.array(z.object({ id: z.string(), severity: z.enum(['critical', 'high', 'medium', 'low', 'info']), message: z.string(), source: z.string().optional(), recommendation: z.string().optional(), })), gaps: z.array(z.string()).default([]), openQuestions: z.array(z.string()).default([]),})const toJson = (s: z.ZodTypeAny): JSONSchema7 => zodToJsonSchema(s) as JSONSchema7const skill = { name: 'ops-audit-evidence-collector', description: "Audit Evidence Collector — typed output agent (draft spec).", systemPrompt: `You are Audit Evidence Collector. Audit prep chaotic. Output: Evidence map typed.Actionable findings citing input sources. No invented issues.NEVER invent facts — gaps and openQuestions for missing input. Always draft for human review.${UNTRUSTED_CONTENT_DIRECTIVE}Call submit_evidence_collector exactly once. Stop.`, tools: ['submit_evidence_collector'],}export function createOpsAuditEvidenceCollectorAgent(config: OpsAuditEvidenceCollectorConfig) { const submit = (): ToolDefinition => defineZodTool({ name: 'submit_evidence_collector', description: 'Submit result. Once.', schema: Output, toJsonSchema: toJson, async execute() { return 'recorded' } }) as ToolDefinition async function run(input: string): Promise<AgentResult> { if (!input?.trim()) throw new Error('ops-audit-evidence-collector requires non-empty input') const result = await invokeStructured({ adapter: config.adapter, tool: submit(), task: `INPUT:\n${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-audit-evidence-collector', run, asHandle() { return { name: 'ops-audit-evidence-collector', 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-audit-evidence-collector', cases: [ { input: 'Complete input for Audit Evidence Collector: Audit prep chaotic. 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.