Skip to content

DCP-02 — Intent Declaration & Policy Gating

No silent actions. An agent declares what it intends to do before doing it; a policy layer decides whether to approve, escalate, or block. DCP-02 defines the declared intent and the policy decision.

Normative artifacts

Artifact Description Schema
Intent Declaration of intent: agent, human, action, channel, data classes, estimated impact. schemas/v1/intent.schema.json
PolicyDecision Decision on an intent: approve / escalate / block, risk score, reasons. Optional: require human confirmation. schemas/v1/policy_decision.schema.json
HumanConfirmation (Optional) Explicit human confirmation on an intent: approve or deny. schemas/v1/human_confirmation.schema.json

$id for schemas: https://dcp-ai.org/schemas/v1/<name>.schema.json

Validation

Validate an object against its schema:

dcp validate schemas/v1/intent.schema.json <intent.json>
dcp validate schemas/v1/policy_decision.schema.json <policy_decision.json>
dcp validate schemas/v1/human_confirmation.schema.json <human_confirmation.json>

A Citizenship Bundle includes Intent and PolicyDecision; validate it with:

dcp validate-bundle <citizenship_bundle.json>

Examples

The source of truth for fields and enums is the JSON Schema; this document is the normative specification that references them.