Conversation
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
![cursor[bot]](https://avatars.githubusercontent.com/in/1210556?s=60&v=4){kind=small}
There was a problem hiding this comment.
Cursor Bugbot has reviewed your changes and found 1 potential issue.
Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.
|
|
||
| // Loop through each message in the plan and send it to the AI | ||
| while (!isAgentFinished && agentExecutionDepth <= AGENT_EXECUTION_DEPTH_LIMIT) { | ||
| while (!isAgentFinished) { |
There was a problem hiding this comment.
Removing agent depth limit risks unbounded execution loop
Medium Severity
The AGENT_EXECUTION_DEPTH_LIMIT constant (previously 20) and its guard in the while loop have been completely removed, leaving while (!isAgentFinished) with no upper bound. If the LLM never returns a finished_task response — due to a model error, looping behavior, or the new Excel-to-Python workflow getting stuck — the agent will make unbounded API calls and run code cells indefinitely with no automatic safety net. Only a manual user stop via shouldContinueAgentExecution can break the loop in that scenario.
2 participants
Description
Gemini 3.1 Pro!
Testing
Try it out
Note
Medium Risk
Removes the agent execution depth limit in the JupyterLab extension, which could allow longer-running or stuck agent loops until explicitly stopped. Other changes are mostly prompt/content and model-name updates with low functional risk.
Overview
Adds Gemini 3.1 Pro support end-to-end by replacing
gemini-3-pro-previewwithgemini-3.1-pro-previewin backend model lists/order (model_utils.py), frontend model constants (models.ts), and associated tests.Enhances agent prompting for Excel conversions by introducing
EXCEL_TO_PYTHON_RULES(newexcel_to_python_rules.py) and injecting it into the agent system message viaprompt_constants.py/agent_system_message.py.Changes agent runtime behavior by removing the
AGENT_EXECUTION_DEPTH_LIMITinuseAgentExecution.ts, so executions no longer auto-stop with a “review and continue” message after a fixed number of steps.Written by Cursor Bugbot for commit b873a83. This will update automatically on new commits. Configure here.