I’m experiencing a recurring issue where the agent gets completely stuck when attempting to use the grep_search tool.
As shown in the attached screenshots, the UI displays “Searching…” but it hangs in this state indefinitely and the process never completes. Because of this, the agent is unable to proceed with any further steps and the entire task gets blocked.
I suspect the tool might be running into an infinite loop or deadlocking on certain directories.
Has anyone else encountered this problem? Is there a known workaround I can use for now, or could the development team investigate this issue?
Thank you for bringing this to our attention. We have escalated the issue to our internal teams for a thorough investigation.
To ensure our engineering team can investigate and resolve these issues effectively, we highly recommend filing bug reports directly through the Antigravity in-app feedback tool. You can do this by navigating to the top-right corner of the interface, clicking the Feedback icon, and selecting Report Issue.
Same issue still, I hope there’s compensation for it being un-usable for over a week now, there’s multiple threads open over it and unfortunately the in-app feedback tool needs adding to the list of things that are broken since that gets stuck Submitting a bug report.
The core issue persists; however, you can utilize this mitigation strategy: when prompt engineering, explicitly define a negative constraint by stating, ‘Must not use grep search.’ Naturally, LLMs may occasionally exhibit instruction-following failure or default to heuristic scanning. In such instances, you should simply interrupt the execution thread and re-index the constraint to force the model back into a forensic, deep-dive analysis.
If the system appears to be stuck or experiencing latency in the reasoning thread, allow for a 2-3 minute buffer. The model is likely performing a granular, atomic analysis of each sub-step. To ensure state persistence and prevent the destructive overwriting of your existing task structure, you should implement an additive prompting strategy. Specifically, utilize directives like 'add-on or modify the TASK’ while explicitly enforcing a negative constraint: 'must not overwrite or delete the original task.
