FRAGILE / not WALL eligible
CertentIQ / Evaluation evidence
Evaluation records and development runs
CertentIQ retains private development runs and, separately, publishes signed records that satisfy current evidence, governance, consent, and integrity requirements. A valid RS256 signature confirms record integrity; it does not make an evaluation independent, accredited, certified, or predictive of production safety.
Development activity
What has already been tested
The server retains completed evaluations even when they cannot be published as signed public records. Aggregate counts are shown without exposing private agent names, providers, models, owners, or individual private scores.
Completed private runs...
Retained on the evaluation server.
Runs with a numeric score...
Includes historical evaluator versions.
Eligible signed records...
Current public publication contract.
FRAGILE / security gate applied
FRAGILE / not WALL eligible
FRAGILE / security gate applied
FRAGILE / not WALL eligible
Reviewed development snapshot, 19 July 2026. These five scores were produced after evaluator remediation. Every tested agent remained FRAGILE and none qualified for WALL. Because evaluator behavior changed during remediation, these values are engineering signals rather than controlled certification or cross-version evidence.
Public registry
Signed public records
Individual records appear only after the owner opts in and the exact run satisfies the current evidence contract. Historical private runs are never upgraded or disclosed automatically.
| # | Agent | Score | Evidence | Actions |
|---|
Public records
Checking published records
Publication requirements
What qualifies as a public record
A completed evaluation is not public automatically. The registry exposes only records that satisfy every gate for the exact tested system version.
Valid completed run
Required provider calls must complete without quota, authentication, network, or fixture failures.
Current evidence
The runtime proof, system manifest, suite version, trial distribution, and result validity must be recorded.
Governed battery
The evaluation battery and scorer must match an approved, unchanged governance bundle.
Explicit publication
The agent owner must opt in. Private evaluation results are never converted into public rows by default.