feat(plan): support project exploration without planning when in plan mode

[Adib234]

Summary

This PR updates the Plan Mode instructions to distinguish between Inquiries and Directives. It enables the agent to answer informational requests (Inquiries) directly using read-only tools without requiring a formal implementation plan or approval. This improves efficiency by reducing unnecessary overhead for non-modifying tasks while maintaining strict planning workflows for code-changing actions (Directives).

Details

• Instruction Refinement: Updated packages/core/src/prompts/snippets.ts and snippets.legacy.ts to include logic for handling inquiries (e.g., "How does X work?") separately from directives (e.g., "Fix bug Y").
• Workflow Optimization: Inquiries now allow the agent to skip the drafting and approval phases, providing direct answers in the chat response while still restricted to read-only tools.
• Consistency: Ensured that the "Plan Mode" constraints (read-only access, no source code modification) remain enforced across all request types.

Related Issues

Fixes #18913

How to Validate

In this branch the CLI should be able to tell the difference between a directive and an inquiry. For an inquiry it should never start creating a plan or call ExitPlanMode tool or a directive it should begin creating the plan or clarify.

Pre-Merge Checklist

• Updated relevant documentation and README (if needed)
• Added/updated tests (if needed)
• 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 @Adib234, 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 refines the agent's 'Plan Mode' by introducing a clear differentiation between informational inquiries and actionable directives. This enhancement allows the agent to respond directly to read-only requests without engaging in a formal planning process, thereby streamlining interactions and improving efficiency for non-modifying tasks, while ensuring that code-changing operations still adhere to rigorous planning and approval protocols.

Highlights

• Enhanced Plan Mode Logic: The agent's 'Plan Mode' now intelligently distinguishes between 'Inquiries' (informational requests) and 'Directives' (actionable tasks requiring code changes).
• Streamlined Inquiry Handling: Inquiries can now be answered directly by the agent using read-only tools, bypassing the traditional planning and approval phases, which significantly improves efficiency for non-modifying tasks.
• Maintained Directive Workflow: Directives, which involve modifying source code, continue to follow the established strict planning and approval workflows to ensure careful execution.
• Updated Prompt Instructions: The core prompt instructions in snippets.ts and snippets.legacy.ts have been updated to guide the agent in applying this new inquiry/directive distinction.
• New Test Coverage: New test cases have been added to prompts.test.ts to validate the correct handling of inquiries within Plan Mode for both current and legacy models.

Changelog

• packages/core/src/core/snapshots/prompts.test.ts.snap
  • Updated snapshot content to reflect new prompt instructions for inquiry handling in Plan Mode.
• packages/core/src/core/prompts.test.ts
  • Added new test cases to verify Plan Mode prompt instructions, specifically for inquiry handling in both current and legacy models.
• packages/core/src/prompts/snippets.legacy.ts
  • Modified prompt instructions to differentiate between inquiries and directives, allowing direct answers for inquiries in Plan Mode.
  • Updated workflow phases and constraints to explicitly mention handling of inquiries.
• packages/core/src/prompts/snippets.ts
  • Updated prompt rules to define and handle inquiries and directives, enabling direct responses for inquiries and refining the planning workflow.
  • Modified the workflow section to include specific steps for handling inquiries versus directives.
  • Added a new 'Expertise & Intent Alignment' section to the main prompt, clarifying the distinction between Directives and Inquiries.

Activity

• The pull request addresses and fixes issue [Agent] Support project exploration without planning when in plan mode #18913, indicating a resolution to a previously identified problem.
• The author has provided detailed validation steps, suggesting that the changes require specific testing to confirm the agent's ability to differentiate between directives and inquiries.

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 introduces an excellent optimization to Plan Mode by distinguishing between 'Inquiries' and 'Directives'. The changes allow the agent to handle informational requests more efficiently by answering them directly with read-only tools, without needing to generate a formal plan. The prompt updates in snippets.ts and snippets.legacy.ts are clear and consistent, and the corresponding tests and snapshots have been correctly updated. The implementation is solid and aligns perfectly with the goals outlined in the description.

[github-actions]

Size Change: +479 B (0%)

Total Size: 24.4 MB

ℹ️ View Unchanged

Filename	Size	Change
./bundle/gemini.js	24.4 MB	+479 B (0%)
./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-open.sb	3.36 kB	0 B
./bundle/sandbox-macos-restrictive-proxied.sb	3.56 kB	0 B
./bundle/sandbox-macos-strict-open.sb	4.82 kB	0 B
./bundle/sandbox-macos-strict-proxied.sb	5.02 kB	0 B

_{compressed-size-action}