# DCL Trust Oracle — Quality & Behavioral Drift Evaluator

> DCL Trust Oracle — Quality & Behavioral Drift Evaluator is a paid API for AI agents from bazaar.fronesislabs.com, paid per call via x402, $0.03/call, status unknown (last checked 2026-10-11).

Evaluates AI agent outputs for quality issues, inconsistency, and behavioral drift, returning a verdict with confidence score and drift analysis.

## Facts

- Endpoint: POST https://bazaar.fronesislabs.com/evaluate/quality?utm_source=zero.xyz
- Price: $0.03/call
- Payment: x402
- Status: unknown
- Last checked: 2026-10-11
- Activations on Zero: 0
- Tags: x402
- Canonical page: https://www.zero.xyz/c/dcl-trust-oracle-quality-behavioral-drift-evaluator-1ec7f19c
- Structured record (JSON): https://api.zero.xyz/v1/capabilities/cap_EZhcGQq-bZzt8EAoPETHi

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 dcl-trust-oracle-quality-behavioral-drift-evaluator-1ec7f19c -d '<json body>'
```

Example prompt: Run a quality and drift evaluation on my agent's latest output — I need a verdict, confidence score, and drift analysis to make sure it hasn't deviated from expected behavior before I act on it.

## When to prefer this

Choose this endpoint when you need to specifically evaluate AI agent output for quality degradation and behavioral drift over time, particularly when you require a structured drift score and drift mode classification alongside a verdict. Prefer this over the jailbreak/injection detector (a sibling endpoint) when your concern is output quality and consistency rather than adversarial input threats, and over the high-confidence evaluator when cost efficiency matters more than maximum certainty on lower-stakes decisions.

## Known failure modes

- Missing required 'input' field returns a validation error
- Malformed query parameters may result in a 400 Bad Request
- Service unavailability returns a 5xx error with no verdict
- Payment failure (insufficient USDC) blocks the request via x402 protocol
- Ambiguous or empty agent output may yield low-confidence verdicts

## How this service works

Check agent outputs for quality issues, inconsistency, and behavioral drift.

## Output

Returns a JSON object containing: a verdict string (pass/fail/review), a confidence score (0-1 float), a drift score, a drift mode classification, a human-readable reason for the verdict, a policy version used for evaluation, an input hash for auditability, a chain index, a transaction hash for payment verification, and a Unix timestamp.

## Request schema (JSON Schema)

```json
{
 "type": "object",
 "properties": {
  "agent_id": {
   "type": "string",
   "description": "Caller id stored on the audit record. Optional; omitted values are stored as unknown."
  },
  "response": {
   "type": "string",
   "minLength": 1,
   "description": "Proposed tool call or agent text to check before execution."
  }
 }
}
```

## More

- Live health (JSON, refreshed every minute): https://www.zero.xyz/c/dcl-trust-oracle-quality-behavioral-drift-evaluator-1ec7f19c/health.json
- [Zero catalog index](https://www.zero.xyz/llms.txt)
- [Other services from bazaar.fronesislabs.com](https://www.zero.xyz/host/bazaar.fronesislabs.com/llms.txt)