Record an agent decision

POST/v1/agent_decisions

Record a human reviewer’s final verdict for content that SafetyKit has previously reviewed. This ground truth feedback allows SafetyKit to improve the accuracy of automated decisions. Each call provides feedback on whether a given piece of content violates a specific policy category. If content is reviewed for multiple policies, send one call per policy.

Body ParametersJSONExpand Collapse

content_type: string

The dataapi namespace of the content being reviewed

decision: string

The human reviewer’s verdict on whether the content violates the specified policy (‘match’/‘no_match’).

policy_id: string

The SafetyKit policy category ID that was evaluated, as configured for your account

reference_id: string

The external id for the content being reviewed

action: optional string

The correct enforcement action for this content (e.g. ‘reduce_visibility’, ‘unpublish’)

comment: optional string

Additional context or details from the human reviewer

ReturnsExpand Collapse

safetykitCorrect: boolean

Whether the SafetyKit decision matched the agent decision

Record an agent decision

curl https://api.safetykit.com/v1/agent_decisions \
    -H 'Content-Type: application/json' \
    -H "Authorization: Bearer $SAFETYKIT_API_KEY" \
    -d '{
          "content_type": "product_listing",
          "decision": "match",
          "policy_id": "counterfeit",
          "reference_id": "listing_12345"
        }'

{
  "safetykitCorrect": true
}

Returns Examples

{
  "safetykitCorrect": true
}