fix(core): prioritize conditional policy rules and harden Plan Mode

[Abhijit-2592]

Summary

This PR ensures that conditional policy rules (using ) are correctly prioritized over global exclusion rules in the . This fixes a critical issue in Plan Mode where authorized tools (like for specific paths) were being blocked by the global catch-all deny. It also clarifies that policy rules without explicit apply to all modes by default for safety.

Details

• Prioritize Conditional Rules: Updated to ensure that high-priority rules with are marked as processed. This prevents lower-priority global DENY rules (like the catch-all in Plan Mode) from overriding specific allowances.
• Implicit Mode Inheritance: Rules without an explicit array now correctly apply to all modes. This prevents accidental security gaps where a general deny rule might be ignored in specific modes if not explicitly listed.
• Plan Mode Verification: Verified that is correctly denied in Plan Mode by the global catch-all, while ensuring authorized file editing tools function as intended.
• Cleanup: Removed redundant explicit deny rules in that are now safely covered by the improved engine logic.

Related Issues

Closes #18881

How to Validate

1. Run core unit tests (including new Plan Mode simulation):

   npm test -w @google/gemini-cli-core -- src/policy/policy-engine.test.ts

2. Run full preflight suite:

   npm run preflight

Pre-Merge Checklist

• Updated relevant documentation and README (if needed)
• Added/updated tests (verified Plan Mode behavior)
• Noted breaking changes (if any)
• Validated on required platforms/methods:
  • MacOS
    • npm run
    • npx
    • Docker
    • Podman
    • Seatbelt
  • Windows
    • npm run
    • npx
    • Docker
  • Linux
    • npm run
    • npx
    • Docker

[gemini-code-assist]

Summary of Changes

Hello @Abhijit-2592, 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 strengthens the PolicyEngine's security and predictability by enforcing explicit mode declarations for policy rules. It prevents policy rules from inadvertently affecting unintended operational modes and specifically hardens the Plan mode against arbitrary shell command execution, thereby reducing potential security risks.

Highlights

• Explicit Mode Requirement for Policy Rules: Policy rules now require an explicit 'modes' array to be considered for tool exclusion, preventing unintended rule application across different approval modes.
• Enhanced Plan Mode Security: The 'run_shell_command' tool is now explicitly denied in 'Plan' mode with a high priority, significantly enhancing security by preventing arbitrary shell command execution.
• Expanded Test Coverage: Test coverage for the policy engine has been expanded to validate the new explicit mode requirement and existing tests have been updated to reflect this change.

Changelog

• packages/core/src/policy/policies/plan.toml
  • Added a new policy rule to explicitly deny 'run_shell_command' in 'plan' mode with a priority of 100.
  • Updated comments to clarify the behavior of the catch-all deny rule.
• packages/core/src/policy/policy-engine.test.ts
  • Introduced a new test case to verify that policy rules without an explicit 'modes' array are ignored.
  • Modified existing test cases within 'getExcludedTools' to include explicit 'modes' arrays for policy rules, aligning with the new requirement.
• packages/core/src/policy/policy-engine.ts
  • Updated the 'getExcludedTools' method to skip any policy rule that does not define an explicit 'modes' array, ensuring rules are only applied when their intended modes are specified.

Activity

• The author has marked the 'Added/updated tests' checklist item as complete.
• The author has marked the 'Validated on required platforms/methods: MacOS (npm run)' checklist item as complete.

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 enhances the policy engine's security and clarity by requiring policy rules to explicitly declare modes for tool exclusion. This prevents rules from one mode accidentally affecting another. The changes are well-implemented in PolicyEngine.getExcludedTools and are accompanied by thorough updates to the test suite, including a new test case for the explicit mode requirement. Additionally, a specific security hardening has been added to plan.toml to explicitly deny run_shell_command in plan mode. The changes are logical and correctly implemented. I have reviewed the code and found no issues of high or critical severity.

[github-actions]

Size Change: +133 B (0%)

Total Size: 24.3 MB

ℹ️ View Unchanged

Filename	Size	Change
./bundle/gemini.js	24.3 MB	+133 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}

[Abhijit-2592]

This change ensures that the PLAN mode has the right set of tools as shown in the SC