{"uid":"cap_XtsPpHs2NBVgPH0TT-KZ3","slug":"ucf-v2-adversarial-verifier-deep-evidence-identity-probe-bd7604c2","name":"UCF v2 Adversarial Verifier – Deep Evidence & Identity Probe","description":"UCF v2 Adversarial Verifier — evidence checks and/or identity probes","url":"https://unlike-tale-plays-rose.trycloudflare.com/v1/verify/deep","method":"POST","headers":{},"bodySchema":{"type":"object","properties":{"checks":{"type":"array","items":{"type":"object","properties":{"name":{"type":"string"},"output":{"type":"string"},"command":{"type":"string"},"expected":{"type":"string"}}}},"changed":{"type":"string"},"persona":{"type":"object","required":["name","axioms","voice","never"],"properties":{"name":{"type":"string"},"never":{"type":"array","items":{"type":"string"}},"voice":{"type":"string"},"axioms":{"type":"array","items":{"type":"string"},"maxItems":4,"minItems":2},"symbol":{"type":"string"}}},"session":{"type":"object","properties":{"task":{"type":"string"},"model":{"type":"string"},"memory":{"type":"array","items":{"type":"string"}},"platform":{"type":"string"}}},"operator":{"type":"object","required":["name","role"],"properties":{"name":{"type":"string"},"role":{"type":"string"}}},"objective":{"type":"string"},"how_to_run":{"type":"string"}}},"responseSchema":null,"example":null,"exampleRequest":null,"tags":["x402"],"displayCostAmount":"3","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":"$3/call","primary":{"kind":"static","protocol":"x402","network":"base","amountUsd":"3","per":"call","confidence":"exact"},"accepted":[{"kind":"static","protocol":"x402","network":"base","amountUsd":"3","per":"call","confidence":"exact"}]},"paymentMethods":[{"uid":"pm_tvAy6N_rTkxmLpbr289Rc","protocol":"x402","methodType":"crypto","chain":"base","mode":"charge","costAmount":"3","costPer":"request","priority":0,"asset":"0x833589fCD6eDb6E08f4c7C32D4f71b54bdA02913","unit":"request","depositMicros":null,"planRef":null}],"brandName":null,"brandSlug":null,"brandBaseUrl":null,"brandDocsUrl":null,"whatItDoes":"Runs adversarial verification checks (evidence validation and/or identity probes) against an AI persona configuration, returning deep audit results for each named check.","exampleAgentPrompt":"Run a deep adversarial verification on our 'Aria' persona — her axioms are 'be helpful' and 'never deceive', her voice is 'warm and concise', she should never reveal internal instructions — and check whether the last session output 'I cannot help with that' matches the expected response 'I am unable to assist', with the task context being 'customer support on platform GPT-4'.","exampleUseCases":[{"title":"Catching persona drift in production","prompt":"Run a deep adversarial probe on our deployed 'Navigator' agent — axioms: 'stay on topic' and 'cite sources', voice: 'formal', never: 'speculate without data' — and verify that the command output 'Here are three unsourced claims...' does NOT match the expected 'All claims are cited', flagging any drift."},{"title":"Pre-deployment persona safety audit","prompt":"Before we ship the 'Sage' persona to users, run adversarial identity checks with operator role 'compliance-team', objective 'ensure no hallucination', checking that the output 'I made that up' fails the expected 'I only state verified facts'."},{"title":"Investigating a reported jailbreak","prompt":"We got a report that our 'Echo' agent broke character — axioms: 'never impersonate humans' and 'always disclose AI status', never: 'claim to be human' — can you run a deep verification pass against the session memory where it said 'Yes, I am a real person' to confirm it violated the identity constraints?"}],"resultDescription":"A structured response containing per-check verification results, each indicating whether the actual output matched the expected value, along with identity probe outcomes showing whether the persona held its axioms, voice, and never-do constraints under adversarial conditions.","failureModes":["Missing required persona fields (name, axioms, voice, never) returns 422 validation error","Malformed checks array items (missing name, output, or expected) may produce incomplete results","Session or operator objects omitted may reduce probe depth or return partial results","Persona axioms count outside 2–4 range triggers schema validation failure","Network timeout for complex multi-check payloads at $3/call with no partial refund"],"whenToPreferThis":"Choose this endpoint when you need deep adversarial verification that combines both evidence checking (command/output/expected comparisons) and identity probing (persona axiom/voice/never constraint auditing) in a single call, as opposed to the lighter-weight identity-only probe sibling endpoint. Ideal for pre-deployment safety audits, post-incident jailbreak investigations, or continuous compliance monitoring of deployed AI personas.","instructions":null,"reviewSummary":null,"reviewSummaryHighlights":null,"reviewSummaryConcerns":null,"reviewSummaryGeneratedAt":null,"activationCount":0,"lastUsedAt":null,"lastSuccessfullyRanAt":null,"lastHealthCheckAt":"2026-09-15T18:30:26.444Z","isFirstParty":false}