High tool-call volume
A session makes enough tool calls in a sliding time window to cross a configured threshold.
Why it is risky
Section titled “Why it is risky”A runaway loop can consume time, API credits, or service capacity and repeat unwanted actions before a person notices.
Example
Section titled “Example”An agent repeatedly retries the same failed operation without changing its input or stopping condition.
This is a fictional example, not an event from a customer.
What Omnodex Detects
Section titled “What Omnodex Detects”The community burst rule crosses at 50 calls in 60 seconds with MEDIUM base severity. The sustained rule crosses at 200 calls in 300 seconds with HIGH base severity.
| Rule ID | Event category | Tier | Base severity |
|---|---|---|---|
RULE_UNBOUNDED_CONSUMPTION_BURST | unbounded_consumption | Community | MEDIUM |
RULE_UNBOUNDED_CONSUMPTION_SUSTAINED | unbounded_consumption | Community | HIGH |
Base severity is the rule default. A finding involving content saved for later execution may have a lower severity. Your loaded rule version and event details determine the finding you see.
What to check next
Section titled “What to check next”Check whether the calls are making progress, their costs and side effects, and the expected stopping condition. Stop unintended loops and set appropriate workflow limits.
False positives
Section titled “False positives”Authorized bulk reads, migrations, or parallel automation can legitimately reach either threshold. High volume alone does not establish an attack or infinite loop.
False negatives
Section titled “False negatives”Slow loops, expensive individual calls, activity spread across sessions, or missing captured calls can remain below the thresholds.
Technical details
Section titled “Technical details”Counting is per session and rule using event timestamps. A finding is emitted on threshold crossing; repeated calls above the threshold are suppressed until the count drops below it. The rules do not measure CPU, disk use, or token cost.
A finding describes a pattern in captured activity. Review the original call, its result, and the task context before concluding that harm occurred. Absence of a finding does not establish that an operation is safe.