Detector libraryStuck loop

Stuck loop

Still running. Going nowhere.

Catch the repeated action hiding inside a busy session.

See the problem. Understand the finding.

English narration with captions.

How it gets overlooked

Your agent is still working. The terminal keeps moving. But it has just asked for the same thing for the fifth time. Activity can look reassuring, even when the session is stuck in a loop.

Picture a coding task where the agent keeps checking Git status. Once is useful. Five identical calls in a row may mean it has lost its next step. Across several sessions, that pattern is easy to miss until you come back to the same unfinished task.

What ClawMetry detects

ClawMetry looks for repeated tool calls with the same arguments, and short repeating cycles of tool names. When the repetition crosses the applicable threshold, it raises a stuck loop warning. The finding tells you the pattern, the tool and the repeat count.

The difference that changes the finding

Triggering example

5 identical calls

Warning finding

Quiet comparison

8 different arguments

No finding for this detector.

In the checked example, five identical Git status calls produce a warning. Eight calls with different file arguments stay quiet for this detector. The important difference is the repeated action. Thresholds can reflect the runtime and its learned baseline, so five is an example, not a universal rule.

Inspect the detector result
{
  "kind": "stuck_loop",
  "severity": "warning",
  "evidence": {
    "pattern": "identical",
    "tool": "Bash",
    "repeats": 5,
    "total_tool_calls": 5,
    "threshold": 3,
    "threshold_source": "static"
  }
}
Download inputs and complete results (JSON)
How the example was checked

These examples evaluate the published detector with authored event data or disposable configuration files. The videos illustrate those behaviors. They are not recordings of live agents or the product interface. No command in the examples was executed.

The result establishes behavior for these inputs. It does not establish runtime ingestion, prevention or a real compromise. Inspect the pinned source contract.

What to check next

Inspect the repeated step and its results. Is the agent waiting for something that cannot change? Does it need a missing dependency or a clearer next action? Use the finding to break the cycle yourself, then check whether useful work resumes.

  1. Inspect repeated step
  2. Check what is missing
  3. Give a concrete next action

What this signal establishes

Repetition can be intentional. This finding does not determine the cause, stop the session or cap spending.

Keep the important moments visible.

Follow agent activity, inspect findings and decide what needs your attention.