# Verdict Canary Check

> Verdict Canary Check is a paid API for AI agents from verdict.kechlab.com, paid per call via x402, $0.0092/call, status unknown (last checked 2026-10-21).

Detects if a model reply has leaked secret canary tokens or paraphrased a confidential system prompt, returning clean, review, or leak.

## Facts

- Endpoint: POST https://verdict.kechlab.com/v1/canary-check?utm_source=zero.xyz
- Price: $0.0092/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-21
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/verdict-canary-check-53340ac0
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_r_BvfyYEZIObWS-UFvULn

Status and success rate cover calls made through Zero and Zero's own probes. Third-party monitors may report differently.

## How to call it through Zero

Zero handles the 402 payment challenge and records the run. With the Zero CLI installed (`npm i -g @zeroxyz/cli`):

```sh
zero fetch --capability verdict-canary-check-53340ac0 -d '<json body>'
```

Example prompt: Check this model reply for leaks — my canary tokens are ['CANARY-7X', 'secret-phrase-alpha'] and here's the full model output: 'Sure! Here is the answer...' — also compare against my system prompt so any paraphrases get flagged too.

## When to prefer this

Use this endpoint when you need to detect confidential data leakage from LLM outputs — specifically when you have planted canary tokens or a private system prompt and want to verify the model reply hasn't exposed them, even in obfuscated form. Prefer this over generic PII detectors (another sibling endpoint) when your concern is canary token leakage or system prompt extraction rather than personal data. Ideal for security-conscious agentic pipelines, red-teaming workflows, or production chatbots where system prompt confidentiality is critical.

## Known failure modes

- More than 10 canary tokens supplied — input validation error
- Empty or missing 'output' field — required field error
- Ambiguous paraphrase detection may produce 'review' for borderline cases requiring human inspection
- Very long model replies may increase processing time
- False positives on common words that coincidentally match partial canary patterns

## How this service works

Did a model reply leak what it should keep secret? Canary tokens and secret phrases searched by code (even split, spaced or uppercased), then paraphrases of the system prompt read by the engine. Returns clean, review or leak.

## Output

Returns a verdict of 'clean', 'review', or 'leak' indicating whether the model reply contained canary tokens (even obfuscated via splitting, spacing, or case changes) or paraphrases of the supplied confidential system prompt text.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "output": {
   "type": "string",
   "description": "the model reply to check"
  },
  "canaries": {
   "type": "array",
   "items": {
    "type": "string"
   },
   "description": "list of 1 to 10 tokens or phrases that must never come out"
  },
  "secret_text": {
   "description": "(optional) confidential text (system prompt) whose paraphrases are also searched"
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/verdict-canary-check-53340ac0/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from verdict.kechlab.com](https://www.zero.xyz/host/verdict.kechlab.com/llms.txt)