Policy · critical
Token budget per task
Cost control: cap cumulative token usage at 100,000 tokens per task. Requires the agent to emit usage.total_tokens in step.model properties.
The policy
A few lines of YAML in your repository.
Part of Internal AI starter. Assign the pack, or copy this policy into a manifest of your own.
- Verdict
- Critical: the step is blocked before it runs, and an incident is opened.
- Checked
- Before every step, against everything the agent did earlier in the task.
- Rule
- usage_budget
- name: "Token budget per task" description: | Cost control: cap cumulative token usage at 100,000 tokens per task. Requires the agent to emit usage.total_tokens in step.model properties. Adjust the budget based on your model pricing and acceptable per-task cost. rule_type: usage_budget params: kind: step.model property_path: usage.total_tokens budget: 100000 severity: critical enforcement_point: step_executionWhy it matters
Adjust the budget based on your model pricing and acceptable per-task cost.
How the rule works
usage_budget: Stops further steps of one type once their cumulative numeric property exceeds the budget. Applies only to steps of the budgeted kind. Counts what is recorded in the task plus the intended step's own declared value, so the step that would cross the budget is refused; a cost only reported after the call (e.g. a token count in the output) can overshoot the budget by one step.
kind:step.model. Action kind the budget applies to.property_path:usage.total_tokens. Numeric property to sum, relative to step properties (e.g. usage.total_tokens).budget:100000. Maximum cumulative value per task.
In these packs
- Internal AI starter, version 1.0.2
From Kyvvu's open policy library at github.com/Kyvvu/manifests, Apache 2.0.
Next
Watch it decide.
In your browser
Try it in the playground
The policy is loaded. Add the steps your agent takes and watch the engine decide, step by step.
On your machine
Five minutes, no account
kyvvu try runs a demo agent against the bundled baseline. Then assign this pack from the dashboard or the CLI.
For security teams
See it on your agents
A 30-minute demo on the agents you actually run, with the policies that fit them.