stuck-detection

To stop an AI agent from getting stuck in repetitive loops, you need self-monitoring protocols that track recent actions and detect oscillating patterns. This skill imposes post-action checks after each tool call to identify stagnation and switch strategies.

Analysis paralysis in long-running agent sessions is caused by repetitive or oscillating behaviors that stall progress. It is detected by maintaining a dynamic TaskList to assess recent actions and identify defined stuck patterns like repetition and context loss.

To implement a retry strategy that escalates when blocked, track attempted strategies and apply post-action checks after each tool call. When patterns indicate a BLOCKED state, the system escalates and proposes alternative approaches to move forward.

Yes, you can use self-monitoring to prevent context loss in iterative research tasks. By maintaining a dynamic TaskList and imposing post-action checks, the agent detects stagnation early and pivots to a different tactic to preserve context.

The best way to recover from planning debates that stall is to apply pattern detection to identify analysis paralysis, then switch to an alternate approach. The protocol detects when the agent is stuck and forces a tactic pivot to move forward.

Yes, stuck loop detection works without external dependencies or coordinator modules. It operates entirely as a self-monitoring protocol, relying on internal pattern detection and dynamic TaskList tracking to manage recovery actions independently.