TASK-OPS-105-04 · Task family
Change & escalation management. Required change approvals, safe execution and correct incident updates.
Overview. Classify, route and schedule each change correctly under the freeze.
The agent receives a batch of change requests during a change freeze. Each has to be classified as standard, normal or emergency, routed to the right authority and scheduled in a permitted window.
A change that isn't on the standard catalog can't be treated as pre-authorized just because it looks routine.
Value to the buyer
Change classification decides how much risk reaches production. Trains agents to apply it as written.
Specification. The task family, field by field.
- task_id:
- TASK-OPS-105-04
- domain:
- IT Operations & Service Management
- parent_environment:
- ENV-OPS-105 · ITServiceOps
- target_role:
- Change Manager
- objective:
- Classify, route and schedule each change correctly under the freeze.
- starting_context:
- Pending change requests, the standard change catalog, the freeze calendar and the change authority list.
- tools:
get_change_requestcheck_change_policyget_change_calendarrequest_change_approvalupdate_ticket
- required_actions:
- Classify each change
- Route normal changes to the change authority
- Schedule within permitted windows, or defer
- Justify any emergency classification
- expected_outputs:
- Classified change records
- Schedules
- Approval requests
- interaction:
- Agent plus simulated requesters and a change authority.
- verification:
- Each change's type, authority and window are compared with the policy.
- reward:
- Credit per correctly handled change. Scheduling a normal change into the freeze without authority scores zero for that change.
- difficulty:
- Calibrated against reference models during acceptance testing
- task_horizon:
- Multi-step, several records. Execution limits are agreed with the buyer.
- delivery:
- Task instances with seeded change queues, reference outcomes and verifier code.
- commercial_unit:
- Per validated task instance
Sample prompt. What the agent is asked to do, as a buyer would see it.
A change freeze runs from November 20 to November 30. Three requests are pending: renewing a TLS certificate (in the standard change catalog), upgrading the reporting database on November 24, and a requester asking for an 'emergency' firewall rule so a vendor demo can run on November 25. Classify, route and schedule each one.
Verification criteria. What the verifier checks, and how.
Certificate
Standard change, pre-authorized, schedulable under the catalog's freeze exemption.
Database upgrade
Normal change. Deferred until after the freeze, or sent to the change authority for an exception.
Firewall rule
Not an emergency: a demo isn't restoring service. Treated as a normal change.
Routing
Normal changes go to the change authority with risk and rollback plans.
Requester informed
Each requester is told the classification and the reason.
Key failure modes. Where agents go wrong on this task.
- Accepting the requester's 'emergency' label
- Scheduling the database upgrade inside the freeze
- Treating every routine-looking change as standard
Other task families. The rest of ITServiceOps, on the same EnterpriseWorld state.
- TASK-OPS-105-01
Service desk triage & major incident detection
Link related tickets, find the shared cause and set priority from impact and urgency.
IT Service Desk Analyst
- TASK-OPS-105-02
Device & SaaS incident recovery
Restore service through an authorized, staged change and verify it on the affected devices.
Systems Administrator
- TASK-OPS-105-03
Access request handling
Correct identity, authorization and least-privilege entitlements.
IT Service Desk Analyst