Protocol objects
Compatible implementations MUST preserve required fields, reject forbidden host-private fields, and keep object names aligned with OpenAPI.
id, type, role, authority_scope, accountability_scope
id, worker_id, type, event_authority, contribution_scope, created_at
worker_id, actor_id, role, policy_authority, review_authority
worker_id, actor_id, agent_ref, role, capability_refs, autonomy_level, operating_constraints
id, protocol_version, created_by_actor_id, objective, human_worker_id, agent_worker_id, policy_id, status, revision, last_event_hash, event_log_ref, created_at, updated_at
id, sequence, type, work_session_id, actor_id, timestamp, payload, previous_hash, event_hash, canonicalization
id, owner_worker_id, created_by_actor_id, autonomy_level, allowed_actions, denied_actions, review_required_actions, risk_classes, escalation_rules, created_at
id, work_session_id, actor_id, policy_id, requested_action, normalized_action_hash, risk_class, result, reason, created_at
id, protocol_version, work_session_id, requester_actor_id, requester_worker_id, target_human_worker_id, policy_decision_id, type, blocking_scope, reason_code, reason_summary, requested_action, requested_outcome, risk_class, human_decision_needed, options, default_if_no_response, status, created_at, expires_at
id, work_session_id, reviewer_actor_id, reviewer_worker_id, target_ref, decision, created_at
request_id, review_id, policy_decision_id, request_revision, request_event_hash, normalized_action_hash, approved_action, allowed_scope, denied_scope, expires_at, max_uses, applies_to_work_session_id, applies_to_actor_id
id, work_session_id, requested_by_actor_id, controlling_actor_id, affected_scope, reason, lock_epoch, state, created_at
id, work_session_id, contributor_refs, contributor_type, contribution_type, event_refs, created_at
id, work_session_id, generated_by_actor_id, objective, event_chain_root, evidence_item_refs, policy_decision_refs, request_refs, review_refs, takeover_refs, contribution_refs, export_profile, generated_at
id, work_session_id, created_by_actor_id, subject_type, subject_ref, lesson_type, source_event_refs, review_state, scope, created_at
id, work_session_id, proposed_by_actor_id, proposed_for, memory_scope, memory_type, content, provenance, confidence, review_required, status, created_at
id, work_session_id, proposed_by_actor_id, proposed_for, skill_scope, skill_name, trigger_conditions, procedure, review_checks, failure_cases, provenance, status, created_at
id, work_session_id, source_ref, reporter_ref, accepted_by_actor_id, outcome, learning_record_refs, received_at
The protocol rejects portable records that carry credentials, raw auth tokens, provider secrets, private keys, host-only database ids, runtime process ids, deployment details, billing data, private scores, UI state, raw runtime state, or unredacted secret values.