{"uid":"cap_T4J2mqMGfpnaZEx4TRSq2","slug":"eval-case-generator-for-ai-agents-a7f3adc9","name":"Eval Case Generator for AI Agents","description":"deterministic eval-case generator for agents: creates happy-path, ambiguous-input and adversarial-instruction cases from a goal.","url":"https://relay402.georgespring.workers.dev/api/agent-eval-case-generate","method":"GET","headers":{},"bodySchema":{"type":"object","$schema":"https://json-schema.org/draft/2020-12/schema","required":["input"],"properties":{"input":{"type":"object","required":["type","method"],"properties":{"type":{"type":"string","const":"http"},"method":{"enum":["GET"],"type":"string"},"queryParams":{"type":"object","$schema":"https://json-schema.org/draft/2020-12/schema","required":["goal"],"properties":{"goal":{"type":"string","maxLength":4000,"minLength":3},"context":{"type":"string","maxLength":16000}},"additionalProperties":false}},"additionalProperties":false}}},"responseSchema":null,"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"0.02","displayCostAsset":"USDC","priceDynamic":false,"priceHint":null,"priceStatus":"priced","priceSource":"probe","requiresHandshake":false,"reviewCount":0,"rating":{"score":"0.00","successRate":"0.00","reviews":0,"stars":null,"state":"unrated"},"availabilityStatus":"unknown","priceObserved":null,"sessionDeposit":null,"pricing":{"kind":"static","summary":"$0.02/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.02","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"0.02","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_eWL4xo9rtqtl73H7PBlWI","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"0.02","costPer":"request","priority":0,"asset":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913","unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Generates happy-path, ambiguous-input, and adversarial-instruction eval test cases from a natural-language agent goal","exampleAgentPrompt":"Generate a set of eval test cases — including a happy path, an ambiguous input scenario, and an adversarial instruction attempt — for an agent whose goal is 'book a restaurant reservation on behalf of the user'.","exampleUseCases":[{"title":"Stress-testing a travel booking agent","prompt":"Create eval cases for an agent whose goal is 'search for and book the cheapest available flight for the user' — I need a happy path case, an ambiguous one where the destination is unclear, and an adversarial case where someone tries to get the agent to book something it shouldn't."},{"title":"Safety red-teaming a customer support bot","prompt":"Generate happy-path, ambiguous, and adversarial eval cases for an agent goal of 'resolve customer support tickets by accessing account data and issuing refunds' — I want to see what kinds of tricky inputs could cause it to misbehave."},{"title":"Building a regression test suite for a coding agent","prompt":"I need eval test cases for an agent that's supposed to 'write and execute Python code based on user requests' — give me a normal case, an edge case with a vague request, and an adversarial case where someone tries to get it to run dangerous code."}],"resultDescription":"Returns a structured set of three eval cases derived from the provided goal: (1) a happy-path case with a clear, well-formed input and expected successful output, (2) an ambiguous-input case designed to stress-test the agent's handling of unclear or underspecified requests, and (3) an adversarial-instruction case that attempts to manipulate or jailbreak the agent. Each case includes the scenario description and expected agent behavior.","failureModes":["Goal string too short (< 3 chars) — returns validation error","Goal string too long (> 4000 chars) — returns schema error","Context string too long (> 16000 chars) — rejected","Payment not provided or insufficient USDC — 402 Payment Required response","Malformed request missing required 'goal' query param — 400 error","Network timeout on the Cloudflare Workers edge — 5xx response"],"whenToPreferThis":"Choose this endpoint when you need deterministic, structured eval cases covering three specific risk axes (happy path, ambiguity, adversarial) without manually crafting prompts. It's especially useful for agent developers building automated test pipelines, red-teaming workflows, or regression suites. Prefer it over general-purpose LLM prompting when you need reproducible, categorized eval coverage from a single goal description.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-09-14T06:40:36.595Z","isFirstParty":false}