Unusual Tool Call Velocity in Session
This detects sessions where an AI agent's tools are invoked far more often than normal, 20 or more times in a single session. A high call count can mean someone is scripting an attack or probing the agent's tools automatically rather than using it interactively.
How the attack works
An attacker (or automated script) interacts with an AI agent and repeatedly triggers tool calls in rapid succession within one session. This pattern fits automated probing, where a script tries many tool invocations quickly to map capabilities, find weaknesses, or extract data at scale. The rule simply counts tool-call activity per session and flags sessions that cross the 20-call threshold, without judging what the calls actually did.
Netzilo detection
Netzilo reports this behaviour when it is observed.
- Signature ID
- d2c6956f-7960-5f83-9886-15cab99fcaab
- Severity
- Medium
Why it matters
If real, this behavior lets an attacker rapidly explore or abuse an agent's available tools, which can lead to data exfiltration, resource abuse, or discovery of exploitable tool combinations, at a pace normal human use would not reach.
What you can do
- →Review flagged sessions to see which tools were called and whether the sequence makes sense for the task.
- →Set per-session or per-user rate limits on tool calls where the platform allows it.
- →Exclude known batch-processing or automated test accounts from alerting to reduce noise.
- →Log tool-call arguments and results, not just counts, so flagged sessions can be investigated in detail.
Known benign look-alikes
- Legitimate batch operations
- Automated testing sessions