Record Agent Skill Evaluation | Zep Documentation

from zep_cloud import Zep
client = Zep(
    api_key="YOUR_API_KEY_HERE",
)
client.agent.skill.evaluation.create(
    agent_uuid="agent_uuid",
    skill_uuid="skill_uuid",
    evaluated_by="customer",
    evaluator_version="v1.0.0",
    skill_version=1,
    verdict="succeeded",
)

{
  "controlled_test": false,
  "created_at": "2026-01-01T00:00:00Z",
  "environment": "production",
  "evaluated_by": "customer",
  "evaluator_version": "v1.0.0",
  "evidence_reference": "https://example.com/evidence/refund",
  "metrics": {},
  "skill_uuid": "6a906e70-50ca-4e43-a874-5fcae02a71fa",
  "skill_version": 1,
  "trace_uuid": "6a906e70-50ca-4e43-a874-5fcae02a71fa",
  "use_uuid": "6a906e70-50ca-4e43-a874-5fcae02a71fa",
  "uuid": "6a906e70-50ca-4e43-a874-5fcae02a71fa",
  "verdict": "succeeded"
}

Record Agent Skill Evaluation. Agent Skills is in closed release. Zep enables Agent Skills only for select accounts. A request from an account without access returns 403 permission_denied. To get access, contact support@getzep.com.

Authentication

AuthorizationApi-Key

Type “Api-Key” followed by a space and the project API key.

Path parameters

agent_uuidstringRequired

Agent UUID

skill_uuidstringRequired

Skill UUID, or the 8-character short_id

Request

Evaluation record

evaluated_byenumRequired

The developer selects the evaluation source. Valid values are customer, tool, offline_corpus, policy.

Allowed values:

evaluator_versionstringRequired<=255 characters

The developer supplies the evaluation rule version. It can contain up to 255 characters.

skill_versionintegerRequired>=1

The developer supplies the Skill version to evaluate. It must be at least 1.

verdictenumRequired

The developer supplies evaluation verdict. Valid values are succeeded, failed, partial, unknown.

Allowed values:

controlled_testbooleanOptional

The developer sets this to true when a controlled test produced the result.

environmentstringOptional<=1024 characters

The developer supplies the environment for this evaluation. At most 1024.

evidence_referencestringOptional<=2048 characters

The developer links this evaluation to an evidence record.

metricsmap from strings to doublesOptional

The developer supplies evaluation metrics.

use_uuidstringOptional

The developer can link this evaluation to a recorded Skill use.

Response

Created

controlled_testbooleanOptional

The developer sets this to true when a controlled test produced the result.

created_atstringOptional

The service returns when Zep created this resource.

environmentstringOptional

The service returns the environment used for this evaluation.

evaluated_bystringOptional

The service returns the evaluation source. Valid values are customer, tool, offline_corpus, and policy.

evaluator_versionstringOptional

The service returns the version of the evaluation rule.

evidence_referencestringOptional

The service returns the link to supporting evidence, when one exists.

metricsmap from strings to doublesOptional

The service returns evaluation metrics.

skill_uuidstringOptional

The service returns the UUID of the Skill.

skill_versionintegerOptional

The service returns the version number of the Skill.

trace_uuidstringOptional

The service returns the UUID of the source Trace.

use_uuidstringOptional

The service returns the UUID of the recorded Skill use, when linked.

uuidstringOptional

The service returns the UUID that identifies this resource.

verdictstringOptional

The service returns the evaluation verdict.

Errors

400

Bad Request Error

401

Unauthorized Error

403

Forbidden Error

404

Not Found Error

409

Conflict Error