feat(core): remove hardcoded policy bypass for local subagents

[abhipatel12]

Remove hard-coded subagent policy. This should be handled by policy engine now

[gemini-code-assist]

Summary of Changes

Hello @abhipatel12, I'm Gemini Code Assist¹! I'm currently reviewing this pull request and will post my feedback shortly. In the meantime, here's a summary to help you and other reviewers quickly get up to speed!

This pull request refactors the subagent execution flow by eliminating a hardcoded policy bypass for local subagents. The change centralizes the decision-making process for subagent call confirmations, ensuring that all such decisions are now handled by the designated policy engine. This promotes a more consistent and configurable approach to managing subagent interactions, reducing specific logic within the SubAgentInvocation class itself.

Highlights

• Policy Engine Integration: The hardcoded policy bypass for local subagents has been removed, delegating all subagent execution confirmation logic to the central policy engine.
• Code Simplification: The shouldConfirmExecute method and its associated ToolCallConfirmationDetails import have been removed from the SubAgentInvocation class, simplifying its internal logic.

Changelog

• packages/core/src/agents/subagent-tool.ts
  • Removed the shouldConfirmExecute method from the SubAgentInvocation class.
  • Removed the import of ToolCallConfirmationDetails as it is no longer utilized.

Activity

• No human activity has been recorded on this pull request yet.

Using Gemini Code Assist

The full guide for Gemini Code Assist can be found on our documentation page, here are some quick tips.

Invoking Gemini

You can request assistance from Gemini at any point by creating a comment using either /gemini <command> or @gemini-code-assist <command>. Below is a summary of the supported commands on the current page.

Feature	Command	Description
Code Review	/gemini review	Performs a code review for the current pull request in its current state.
Pull Request Summary	/gemini summary	Provides a summary of the current pull request in its current state.
Comment	@gemini-code-assist	Responds in comments when explicitly tagged, both in pull request comments and review comments.
Help	/gemini help	Displays a list of available commands.

Customization

To customize Gemini Code Assist for GitHub experience, repository maintainers can create a configuration file and/or provide a custom code review style guide (such as PEP-8 for Python) by creating and adding files to a .gemini/ folder in the base of the repository. Detailed instructions can be found here.

Limitations & Feedback

Gemini Code Assist may make mistakes. Please leave feedback on any instances where its feedback is incorrect or counter productive. You can react with 👍 and 👎 on @gemini-code-assist comments. If you're interested in giving your feedback about your experience with Gemini Code Assist for Github and other Google products, sign up here.

You can also get AI-powered code generation, chat, as well as code reviews directly in the IDE at no cost with the Gemini Code Assist IDE Extension.

Footnotes

1. Review the Privacy Notices, Generative AI Prohibited Use Policy, Terms of Service, and learn how to configure Gemini Code Assist in GitHub here. Gemini can make mistakes, so double check it and use code with caution. ↩

[gemini-code-assist]

Code Review

This pull request removes a hardcoded policy bypass for local subagents within SubAgentInvocation, ensuring all local subagent tool calls flow through the BaseToolInvocation logic and are subject to policy checks, which improves consistency. However, this change also removes a delegation mechanism in SubAgentInvocation.shouldConfirmExecute that was crucial for transparent user confirmation prompts, especially with remote agents. This loss of visibility into data sent to external services during confirmation poses a security risk.

I am having trouble creating individual review comments. Click here to see my feedback.

packages/core/src/agents/subagent-tool.ts (88-98)

The removal of the shouldConfirmExecute override in SubAgentInvocation introduces a security regression by reducing the transparency of the user confirmation flow.

Previously, SubAgentInvocation delegated the confirmation check to the underlying RemoteAgentInvocation (via buildSubInvocation), which provided a detailed prompt including the actual query being sent to the remote agent. By removing this override, the tool now falls back to the base BaseToolInvocation.shouldConfirmExecute implementation, which uses SubAgentInvocation.getDescription().

Since SubAgentInvocation.getDescription() only returns a generic string (Delegating to agent '${this.definition.name}'), the user is no longer shown the specific arguments (like the query) being passed to the subagent. This makes it difficult for users to identify and block potentially malicious or sensitive data transfers to external services.

To remediate this, consider restoring the delegation logic while removing the hardcoded bypass for local agents. This ensures that the specific confirmation details of the underlying invocation are used, while still allowing the policy engine to be checked.

[github-actions]

Size Change: -85 B (0%)

Total Size: 23.7 MB

ℹ️ View Unchanged

Filename	Size	Change
./bundle/gemini.js	23.7 MB	-85 B (0%)
./bundle/sandbox-macos-permissive-closed.sb	1.03 kB	0 B
./bundle/sandbox-macos-permissive-open.sb	890 B	0 B
./bundle/sandbox-macos-permissive-proxied.sb	1.31 kB	0 B
./bundle/sandbox-macos-restrictive-closed.sb	3.29 kB	0 B
./bundle/sandbox-macos-restrictive-open.sb	3.36 kB	0 B
./bundle/sandbox-macos-restrictive-proxied.sb	3.56 kB	0 B

_{compressed-size-action}

[adamfweidman]

I think this will break tool confirmations for remote agents.

We should just remove the kind !== remote and jus keep the buildSubInvocation and shouldConfirmExecute, this will automatically delegate to the remote vs local invocations.

[gundermanc]

LGTM with one q.