OUTCOME · Planned · Create or advance
What changes
None today: this route is not executable. Its intended behavior is: when promoted, creates configuration and evidence only. The automation remains DRAFT, admits no trigger, stores no reusable domain credential, and authorizes no underlying action.
WHY IT MATTERS
- Lets people and agents prepare for automation without falsely presenting roadmap scope as a live capability.
- Gives people and agents a contract-backed way to advance automation.
- Makes triggers, execution attempts, callback destinations, delivery evidence, and retry behavior explicit without granting the automation ambient access to another domain.
ISOLATION + AUTHORITY
Tenant, workspace, automation definition, trigger, run, callback endpoint, secret version, delivery attempt, receipt, retry policy, and target-domain authority remain distinct. Creating or running an automation does not bypass the authorization, idempotency, policy, or lifecycle gates of any action it requests. This planning record grants no runtime authority, and only deployed OpenAPI can define an executable public contract.
BEFORE YOU CALL
- First confirm that this operation is marked implemented-contract and exists in the currently deployed OpenAPI document; until then, no production request is valid.
- Its capability-registry profile is provisional integration guidance, not an executable request schema.
- Authenticate at the documented boundary: bearer+scope.
- Treat the proposed name (body), trigger (body), action_profile_id (body), execution_policy (body), step_up_token (body) as planning input only; re-generate the request from deployed OpenAPI before making a call.
- Use one Idempotency-Key only for retries of the same byte-equivalent logical mutation.
WHAT TO DO NEXT
- Keep this operation disabled in clients, agents, SDKs, and workflow automation while it remains planned-contract.
- Use the stated owner, lifecycle, authority boundary, and provisional automation profile to prepare requirements and conformance tests without sending a request.
- Monitor the capability registry for implemented-contract, then re-fetch deployed OpenAPI and validate its exact security, parameters, schemas, responses, and agent metadata before enabling the integration.
AGENT GUIDANCE
- Never call this planned contract, include it in an executable tool registry, or infer runtime availability from this readable page.
- Its capability-registry profile is provisional integration guidance, not an executable request schema.
- Use automation only for the purpose and lifecycle stage described by this operation; do not treat it as authority for an adjacent action.
- Treat the proposed name (body), trigger (body), action_profile_id (body), execution_policy (body), step_up_token (body) as planning input only; re-generate the request from deployed OpenAPI before making a call.
- Distinguish definition state, enablement, trigger admission, run creation, execution outcome, callback delivery, receipt, retry, and the authoritative state of any referenced business resource.
- After a timeout or conflict, read authoritative state before deciding whether an equivalent retry is safe.
- When implementation lands, discard generated requests based on this planning record and rebuild them from the deployed OpenAPI operation.