Guys, what’s going on? Why won’t Gemini Pro let me run a security check on my own code now? Gemini isn’t even that good at doing this. What’s the deal? This is tricky; I need to work and keep my code secure, but I can’t. Please fix this, …
Hello @Erick_Joseph.
With the rollout of Antigravity version 2.1.1 (released June 22, 2026), the engineering team implemented specific agent security fixes and strict mode verifications. Directories such as .vscode and .cache are now flagged as sensitive paths, and the agent will explicitly halt and require user confirmation before accessing them when strict mode is enabled.
To help determine if the system is functioning as intended-
- Can you clarify what specific security prompts, permission requests, or behavioral changes you are observing?
- What exact actions, terminal commands, or workflows reliably trigger this behavior?
Additionally, we recommend utilizing the in-app feedback option (top right icon > Report Issue) to securely submit your system logs directly to our engineering team for deeper diagnostics.
You see, any time you ask Gemini 3.1 Pro to do something, like finding bugs I missed or even finding vulnerabilities in my own code, it refuses to verify it. You have AI to check what you forgot. (I’m getting more and more tempted to switch to Codex.)