Spend approval

Creating new agents

Builds the workflow that watches model spend against a threshold, opens an approval task when it is crossed, and raises the limit by a configured increment only after a human completes that task.

Used when

When a spend limit needs to move on evidence and approval rather than on someone noticing.

Loads first

• The agent-building and batch-workflow skills.
• The continuous-integration skill, for the per-environment deployment path.

Returns

• The repository layout — the workflow's task handlers, the approval webhook handler, shared configuration modules and tests.
• The state-machine shape, from plan through a per-user approval wait to the aggregate, and how the schedule starts it.
• The parameter and secret list, with the exact commands an operator runs per environment.
• The runtime contract: plan output, item schema, task body, webhook filter, ledger schema, and every named rejection reason.
• The permissions actually granted, least-privilege, per component.
• A verification log covering the approval path, a non-approver, a cap rejection, a drifted limit, a replayed webhook, a timeout, and a redrive.
• Rollback notes for both the code and the configuration.