fix(core): enable autonomous plan execution in non-interactive mode by urielefrenvirtusa · Pull Request #29539 · google-gemini/gemini-cli

urielefrenvirtusa · 2026-09-28T17:05:32Z

Summary

Guards synchronous user consultation and agreement waiting directives in Plan Mode prompts with options.interactive. In non-interactive/headless environments (interactive: false), Plan Mode now instructs the agent to synthesize strategy and proceed directly to plan drafting and execution without halting to wait for user confirmation.

Details

In PR #24423, strict consultation requirements were added to Plan Mode prompts, including directives such as STOP and wait, explicit prohibitions against drafting plans in the initial strategy turn, and mandates to reach informal agreement before calling exit_plan_mode. While appropriate for interactive CLI sessions, these instructions caused headless agent runs (e.g. gemini -p "..." -y) to output an alignment message with 0 tool calls. In non-interactive mode, generating a response with 0 tool calls triggers immediate session termination in nonInteractiveCli, leaving the task unexecuted. Furthermore, the prompt instructed the agent to clarify via ask_user, a tool that is excluded and blocked by policy in headless mode.

Changes in this PR:

Updated renderPlanningWorkflow in packages/core/src/prompts/snippets.ts:
- Step 2 (Consult): In interactive mode, retains the full consultation workflow and wait directives. In non-interactive mode, replaces this with Determine Strategy, instructing the model to synthesize findings and proceed directly to Step 3 (Draft) autonomously without halting.
- Step 4 (Review & Approval): In interactive mode, retains the informal agreement requirement. In non-interactive mode, instructs the model to call exit_plan_mode to finalize the plan and begin implementation.
- Rule 3 (Efficiency): Only includes instructions to clarify via ask_user when options.interactive is true.
- Step 3 (Draft): Preserves existing conditional rendering where human Alignment Check is only included when options.interactive is true.
Added comprehensive unit tests in packages/core/src/prompts/planModeNonInteractive.test.ts verifying prompt generation under interactive and non-interactive configurations for both renderPlanningWorkflow and PromptProvider.getCoreSystemPrompt.

Related Issues

Fixes #28913

How to Validate

Run the targeted Plan Mode prompt test suite:

npm test -w @google/gemini-cli-core -- src/prompts/planModeNonInteractive.test.ts

Run all prompt test suites in core:

npm test -w @google/gemini-cli-core -- src/prompts/

Run core snapshot tests to confirm zero regressions in interactive mode:
```
npm test -w @google/gemini-cli-core -- src/core/prompts.test.ts
```

Verify non-interactive CLI tests:

npm test -w @google/gemini-cli -- src/nonInteractiveCli.test.ts

Run lint and typecheck:
```
npm run lint && npm run typecheck
```

Pre-Merge Checklist

Ensure Plan Mode prompts do not mandate synchronous user consultation or block execution turns when running in non-interactive/headless mode. Guards Step 2 consultation, informal agreement exit requirements, and ask_user references with options.interactive to allow autonomous plan drafting and execution. Bug: b/561554615

gemini-code-assist · 2026-09-28T17:05:43Z

Summary of Changes

Hello, 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 enables autonomous operation for the agent in non-interactive (headless) environments by modifying the Plan Mode prompt workflow. Previously, strict consultation and user-agreement directives prevented headless agents from proceeding with tasks. The changes introduce conditional logic to bypass these interactive-only steps, allowing the agent to synthesize strategies and finalize plans autonomously when running in non-interactive mode.

Highlights

Autonomous Plan Execution: Enabled autonomous plan execution in non-interactive mode by removing blocking user consultation and agreement requirements.
Prompt Workflow Refactoring: Updated renderPlanningWorkflow to conditionally render steps based on the interactive flag, ensuring headless agents proceed directly to drafting and implementation.
Testing: Added comprehensive unit tests in planModeNonInteractive.test.ts to verify prompt generation and agent behavior in both interactive and non-interactive configurations.

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 the 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 counterproductive. 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.

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. ↩

github-actions · 2026-09-28T17:05:47Z

📊 PR Size: size/L

Lines changed: 266
Additions: +257
Deletions: -9
Files changed: 3

gemini-code-assist

Code Review

This pull request introduces support for a non-interactive plan mode workflow in the Gemini CLI core prompts. Specifically, it updates the planning workflow rendering in snippets.ts to conditionally display different instructions for Step 2 (Consult vs. Determine Strategy) and Step 4 (Review & Approval) depending on whether the session is interactive. It also refines Rule 3 (Efficiency) to exclude user-interaction directives in non-interactive mode. To support these changes, a new test file planModeNonInteractive.test.ts has been added to validate both interactive and non-interactive prompt generation. I have no feedback to provide as there are no review comments to assess.

urielefrenvirtusa · 2026-09-28T17:20:22Z

/Gemini review

gemini-code-assist

Code Review

This pull request introduces support for a non-interactive workflow in Plan Mode. It updates renderPlanningWorkflow in packages/core/src/prompts/snippets.ts to dynamically adjust the prompt sections (such as 'Consult' vs 'Determine Strategy' and 'Review & Approval') and rules based on whether the interactive option is enabled. Additionally, a comprehensive suite of unit tests has been added in packages/core/src/prompts/planModeNonInteractive.test.ts to verify both interactive and non-interactive prompt generation behaviors. There are no review comments, and I have no feedback to provide.

github-actions · 2026-09-28T19:59:02Z

🚨 Action Required: Eval Regressions Detected

Model: gemini-3-flash-preview

The following trustworthy evaluations passed on main and in recent Nightly runs, but failed in this PR. These regressions must be addressed before merging.

Test Name	Nightly	PR Result	Status
should use automated tools (prettier --write) to fix formatting issues	100%	25%	❌ Regression

The check passed or was cleared for 69 other trustworthy evaluations.

🛠️ Troubleshooting & Fix Instructions

1. Ask Gemini CLI to fix it (Recommended)

Copy and paste this prompt to the agent:

The eval "should use automated tools (prettier --write) to fix formatting issues" in evals/automated-tool-use.eval.ts is failing. Investigate and fix it using the behavioral-evals skill.

2. Reproduce Locally

Run the following command to see the failure trajectory:

GEMINI_MODEL=gemini-3-flash-preview npm run test:all_evals -- evals/automated-tool-use.eval.ts --testNamePattern="should use automated tools (prettier --write) to fix formatting issues"

3. Manual Fix

See the Fixing Guide for detailed troubleshooting steps.

### 🧠 Model Steering Guidance

This PR modifies files that affect the model's behavior (prompts, tools, or instructions).

🚀 Maintainer Reminder: Please ensure that these changes do not regress results on benchmark evals before merging.

This is an automated guidance message triggered by steering logic signatures.

Verify that the agent autonomously drafts an implementation plan and invokes exit_plan_mode without waiting for user agreement when running in non-interactive/headless mode.

github-actions · 2026-09-29T15:39:16Z

🛑 Action Required: Evaluation Approval

Steering changes have been detected in this PR. To prevent regressions, a maintainer must approve the evaluation run before this PR can be merged.

Maintainers:

Go to the Workflow Run Summary.
Click the yellow 'Review deployments' button.
Select the 'eval-gate' environment and click 'Approve'.

Once approved, the evaluation results will be posted here automatically.

urielefrenvirtusa requested a review from a team as a code owner September 28, 2026 17:05

github-actions Bot added the size/m A medium sized PR label Sep 28, 2026

urielefrenvirtusa had a problem deploying to eval-gate September 28, 2026 17:07 — with GitHub Actions Error

gemini-code-assist Bot reviewed Sep 28, 2026

View reviewed changes

Merge branch 'main' into b_561554615

f3b6f3c

urielefrenvirtusa deployed to eval-gate September 28, 2026 17:08 — with GitHub Actions Active

gemini-cli Bot added priority/p1 Important and should be addressed in the near term. area/non-interactive Issues related to GitHub Actions, SDK, 3P Integrations, Shell Scripting, Command line automation labels Sep 28, 2026

gemini-code-assist Bot reviewed Sep 28, 2026

View reviewed changes

DavidAPierce approved these changes Sep 28, 2026

View reviewed changes

test(evals): add behavioral eval for non-interactive plan mode

45ddd4f

Verify that the agent autonomously drafts an implementation plan and invokes exit_plan_mode without waiting for user agreement when running in non-interactive/headless mode.

DavidAPierce deployed to eval-gate September 28, 2026 21:07 — with GitHub Actions Active

github-actions Bot added the size/l A large sized PR label Sep 28, 2026

github-actions Bot mentioned this pull request Sep 29, 2026

📊 AI CLI 工具社区动态日报 2026-09-29 litang9/big_model_radar#468

Open

Merge branch 'main' into b_561554615

a58d398

DavidAPierce added this pull request to the merge queue Sep 29, 2026

DavidAPierce requested a deployment to eval-gate September 29, 2026 15:39 — with GitHub Actions Waiting

Merged via the queue into google-gemini:main with commit f5dc904 Sep 29, 2026
9 of 10 checks passed

urielefrenvirtusa deleted the b_561554615 branch September 29, 2026 16:22

github-actions Bot mentioned this pull request Sep 30, 2026

📊 AI CLI 工具社区动态日报 2026-10-01 litang9/big_model_radar#478

Open

Navigation Menu

Repository navigation

fix(core): enable autonomous plan execution in non-interactive mode - #29539

fix(core): enable autonomous plan execution in non-interactive mode#29539
DavidAPierce merged 4 commits into
google-gemini:mainfrom
urielefrenvirtusa:b_561554615

urielefrenvirtusa commented Sep 28, 2026

Uh oh!

gemini-code-assist Bot commented Sep 28, 2026

Uh oh!

github-actions Bot commented Sep 28, 2026 •

edited

Loading

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

urielefrenvirtusa commented Sep 28, 2026

Uh oh!

gemini-code-assist Bot left a comment

Uh oh!

github-actions Bot commented Sep 28, 2026 •

edited

Loading

1. Ask Gemini CLI to fix it (Recommended)

2. Reproduce Locally

3. Manual Fix

Uh oh!

github-actions Bot commented Sep 29, 2026

Uh oh!

Uh oh!

This branch is waiting to be deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Conversation

urielefrenvirtusa commented Sep 28, 2026

Summary

Details

Related Issues

How to Validate

Pre-Merge Checklist

Uh oh!

gemini-code-assist Bot commented Sep 28, 2026

Summary of Changes

Highlights

Footnotes

Uh oh!

github-actions Bot commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

urielefrenvirtusa commented Sep 28, 2026

Uh oh!

gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

Code Review

Uh oh!

github-actions Bot commented Sep 28, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🚨 Action Required: Eval Regressions Detected

1. Ask Gemini CLI to fix it (Recommended)

2. Reproduce Locally

3. Manual Fix

Uh oh!

github-actions Bot commented Sep 29, 2026

🛑 Action Required: Evaluation Approval

Uh oh!

Uh oh!

This branch is waiting to be deployed

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

github-actions Bot commented Sep 28, 2026 •

edited

Loading

github-actions Bot commented Sep 28, 2026 •

edited

Loading