Skip to content

feat(core): steer model to use edit tool for surgical edits, fix a typo#26480

Merged
aishaneeshah merged 4 commits intomainfrom
fix/issue-24713-surgical-edits
May 5, 2026
Merged

feat(core): steer model to use edit tool for surgical edits, fix a typo#26480
aishaneeshah merged 4 commits intomainfrom
fix/issue-24713-surgical-edits

Conversation

@aishaneeshah
Copy link
Copy Markdown
Contributor

@aishaneeshah aishaneeshah commented May 5, 2026

Summary

This PR updates the tool descriptions for `write_file` and `replace` in the `gemini-3` model family to steer the model towards more efficient and safer editing practices. It specifically encourages using the `replace` tool for surgical edits to existing files to minimize token usage, simplify code reviews, and prevent accidental deletions.

Details

  • Steering in `write_file`: Added a recommendation to use `replace` for targeted edits to existing files.
  • Enhanced `replace` Description: Highlighted that `replace` is preferred for surgical edits as it minimizes token usage, simplifies reviews, and avoids accidental deletions.
  • Prompt Correction: Fixed a typo in the system prompt guidelines (`packages/core/src/prompts/snippets.ts`) where `read_file` was erroneously listed as failing on ambiguous edits instead of `replace`.
  • Snapshot Updates: Updated test snapshots to reflect the improved descriptions.
  • Refinement: Following user steering, existing detail (like context requirements) in the `replace` description for Gemini 3 was preserved.

Related Issues

Fixes #24713, #25568

How to Validate

  1. Run core package unit tests: `npm test -w @google/gemini-cli-core`
  2. Verify that snapshots match the updated descriptions in `packages/core/src/tools/definitions/model-family-sets/gemini-3.ts`.

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
    • Windows
    • Linux
      • npm run

…g files

This change updates tool descriptions for 'write_file' and 'replace' to encourage surgical edits and simplify the 'replace' tool's strict context requirements. It also fixes a typo in the system prompt guidance.
Reverted changes to default-legacy.ts and ensured gemini-3.ts replace description retains all existing details while adding the preference for surgical edits. Updated snapshots accordingly.
@gemini-code-assist
Copy link
Copy Markdown
Contributor

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 improves the guidance provided to the gemini-3 model regarding file editing tools. By clarifying the intended use cases for 'write_file' and 'replace', the changes aim to reduce token consumption, streamline code reviews, and prevent accidental file deletions. Additionally, a minor correction was made to the system prompt guidelines to accurately reflect tool behavior.

Highlights

  • Model Steering: Updated tool descriptions for 'write_file' and 'replace' in the gemini-3 model family to encourage safer, more efficient surgical edits.
  • Prompt Correction: Fixed a typo in the system prompt guidelines where 'read_file' was incorrectly cited as the tool that fails on ambiguous edits instead of 'replace'.
  • Test Updates: Updated relevant test snapshots to reflect the improved tool descriptions.
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.

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.

@github-actions
Copy link
Copy Markdown

github-actions Bot commented May 5, 2026

Size Change: +183 B (0%)

Total Size: 34 MB

Filename Size Change
./bundle/chunk-7U3WSCE6.js 0 B -19.5 kB (removed) 🏆
./bundle/chunk-BOQW3KB6.js 0 B -49.2 kB (removed) 🏆
./bundle/chunk-ECNYAST2.js 0 B -1.97 MB (removed) 🏆
./bundle/chunk-FBZZXD74.js 0 B -14.7 MB (removed) 🏆
./bundle/chunk-GC2OC5H5.js 0 B -3.43 kB (removed) 🏆
./bundle/chunk-L4LYHGDX.js 0 B -658 kB (removed) 🏆
./bundle/chunk-R3LPUZS6.js 0 B -12.5 kB (removed) 🏆
./bundle/chunk-RH2NEQ3L.js 0 B -2.78 MB (removed) 🏆
./bundle/chunk-ZNV6UR3Z.js 0 B -3.8 kB (removed) 🏆
./bundle/core-M55NN2ZJ.js 0 B -48.7 kB (removed) 🏆
./bundle/devtoolsService-KNOWYCCA.js 0 B -28 kB (removed) 🏆
./bundle/gemini-KPVPY5CI.js 0 B -583 kB (removed) 🏆
./bundle/interactiveCli-RTCHLIQM.js 0 B -1.29 MB (removed) 🏆
./bundle/liteRtServerManager-UPWXJHQR.js 0 B -2.11 kB (removed) 🏆
./bundle/oauth2-provider-DZZFRZIX.js 0 B -9.16 kB (removed) 🏆
./bundle/chunk-2HM75NIY.js 2.78 MB +2.78 MB (new file) 🆕
./bundle/chunk-35R4GZ6K.js 14.7 MB +14.7 MB (new file) 🆕
./bundle/chunk-5I4VLEPV.js 658 kB +658 kB (new file) 🆕
./bundle/chunk-KVW7PN3U.js 19.5 kB +19.5 kB (new file) 🆕
./bundle/chunk-M3HQAMED.js 3.43 kB +3.43 kB (new file) 🆕
./bundle/chunk-O7SH2UWI.js 49.2 kB +49.2 kB (new file) 🆕
./bundle/chunk-ONWW7QVQ.js 3.8 kB +3.8 kB (new file) 🆕
./bundle/chunk-VHPOPZ2T.js 1.97 MB +1.97 MB (new file) 🆕
./bundle/chunk-ZDUFCOTW.js 12.5 kB +12.5 kB (new file) 🆕
./bundle/core-S77YZXPO.js 48.7 kB +48.7 kB (new file) 🆕
./bundle/devtoolsService-XXMUF5W5.js 28 kB +28 kB (new file) 🆕
./bundle/gemini-TMZGEMFN.js 583 kB +583 kB (new file) 🆕
./bundle/interactiveCli-63ZJZDZ4.js 1.29 MB +1.29 MB (new file) 🆕
./bundle/liteRtServerManager-UY43RLCX.js 2.11 kB +2.11 kB (new file) 🆕
./bundle/oauth2-provider-REEJB7YO.js 9.16 kB +9.16 kB (new file) 🆕
ℹ️ View Unchanged
Filename Size Change
./bundle/bundled/third_party/index.js 8 MB 0 B
./bundle/chunk-34MYV7JD.js 2.45 kB 0 B
./bundle/chunk-5AUYMPVF.js 858 B 0 B
./bundle/chunk-5PS3AYFU.js 1.18 kB 0 B
./bundle/chunk-664ZODQF.js 124 kB 0 B
./bundle/chunk-DAHVX5MI.js 206 kB 0 B
./bundle/chunk-IUUIT4SU.js 56.5 kB 0 B
./bundle/chunk-RJTRUG2J.js 39.8 kB 0 B
./bundle/cleanup-ORFERCU6.js 0 B -932 B (removed) 🏆
./bundle/devtools-36NN55EP.js 696 kB 0 B
./bundle/dist-T73EYRDX.js 356 B 0 B
./bundle/events-XB7DADIJ.js 418 B 0 B
./bundle/examples/hooks/scripts/on-start.js 188 B 0 B
./bundle/examples/mcp-server/example.js 1.43 kB 0 B
./bundle/gemini.js 5.1 kB 0 B
./bundle/getMachineId-bsd-TXG52NKR.js 1.55 kB 0 B
./bundle/getMachineId-darwin-7OE4DDZ6.js 1.55 kB 0 B
./bundle/getMachineId-linux-SHIFKOOX.js 1.34 kB 0 B
./bundle/getMachineId-unsupported-5U5DOEYY.js 1.06 kB 0 B
./bundle/getMachineId-win-6KLLGOI4.js 1.72 kB 0 B
./bundle/memoryDiscovery-FB7MMKTA.js 0 B -980 B (removed) 🏆
./bundle/multipart-parser-KPBZEGQU.js 11.7 kB 0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/client/main.js 222 kB 0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/_client-assets.js 229 kB 0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/index.js 13.4 kB 0 B
./bundle/node_modules/@google/gemini-cli-devtools/dist/src/types.js 132 B 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-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
./bundle/src-QVCVGIUX.js 47 kB 0 B
./bundle/start-G5BDDDPC.js 0 B -652 B (removed) 🏆
./bundle/tree-sitter-7U6MW5PS.js 274 kB 0 B
./bundle/tree-sitter-bash-34ZGLXVX.js 1.84 MB 0 B
./bundle/cleanup-YGC6OPXE.js 932 B +932 B (new file) 🆕
./bundle/memoryDiscovery-KNBNCO7K.js 980 B +980 B (new file) 🆕
./bundle/start-D3JH2Z2S.js 652 B +652 B (new file) 🆕

compressed-size-action

Copy link
Copy Markdown
Contributor

@gemini-code-assist gemini-code-assist Bot left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Code Review

This pull request updates the prompt guidelines and tool descriptions for the replace and write_file tools. It corrects the prompt text to clarify that the replace tool fails when the old_string is ambiguous and updates tool descriptions to highlight the advantages of using replace for surgical edits, including reduced token usage and simplified reviews. I have no feedback to provide.

@gemini-cli gemini-cli Bot added the area/agent Issues related to Core Agent, Tools, Memory, Sub-Agents, Hooks, Agent Quality label May 5, 2026
@aishaneeshah aishaneeshah changed the title feat(core): steer model to surgical edits and prevent accidental deletions feat(core): steer model to use edit tool for surgical edits, fix a typo May 5, 2026
@aishaneeshah aishaneeshah added this pull request to the merge queue May 5, 2026
@github-merge-queue github-merge-queue Bot removed this pull request from the merge queue due to failed status checks May 5, 2026
@aishaneeshah aishaneeshah added this pull request to the merge queue May 5, 2026
@github-actions
Copy link
Copy Markdown

github-actions Bot commented May 5, 2026

70 tests passed successfully on gemini-3-flash-preview.

🧠 Model Steering Guidance

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

  • ⚠️ Consider adding Evals: No behavioral evaluations (evals/*.eval.ts) were added or updated in this PR. Consider adding a test case to verify the new behavior and prevent regressions.
  • 🚀 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.

@github-merge-queue github-merge-queue Bot removed this pull request from the merge queue due to failed status checks May 5, 2026
@aishaneeshah aishaneeshah added this pull request to the merge queue May 5, 2026
@github-merge-queue github-merge-queue Bot removed this pull request from the merge queue due to failed status checks May 5, 2026
exports[`coreTools snapshots for specific models > Model: gemini-3-pro-preview > snapshot for tool: replace 1`] = `
{
"description": "Replaces text within a file. By default, the tool expects to find and replace exactly ONE occurrence of \`old_string\`. If you want to replace multiple occurrences of the exact same string, set \`allow_multiple\` to true. This tool requires providing significant context around the change to ensure precise targeting.
"description": "Replaces text within a file. By default, the tool expects to find and replace exactly ONE occurrence of \`old_string\`. If you want to replace multiple occurrences of the exact same string, set \`allow_multiple\` to true. This tool is preferred for surgical edits to existing files as it minimizes token usage, simplifies code reviews, and avoids accidental deletions. This tool requires providing significant context around the change to ensure precise targeting.
Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This all looks ok at a glance but perhaps it's worth running evals to check for regressions because in my experience replace is very finicky. I found that changes to edit guidance frequently caused more tool failures or turns.

Replace needs a large enough old_string to not be ambiguous but a small enough one that the model can remember it without typos.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I haven't run evals. The large context requirement was in the existing description - happy to remove it but figured it was something that was tuned earlier.

Addition to steer model to use edit tool was to address a bunch of issues which reported that Gemini CLI uses write_file to edit large files - there by deleting a lot of existing code or other unintentional side effects

In my personal use - I see this all the time specially when updating tests - and need to instruct the model to only edit the test file, or not remove existing comments etc.

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Also - in all of those cases - technically the tool doesn't "fail" but it fails on the overall intent.

@aishaneeshah aishaneeshah added this pull request to the merge queue May 5, 2026
Merged via the queue into main with commit 0218817 May 5, 2026
29 checks passed
@aishaneeshah aishaneeshah deleted the fix/issue-24713-surgical-edits branch May 5, 2026 19:51
kimjune01 pushed a commit to kimjune01/gemini-cli-claude that referenced this pull request May 6, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

area/agent Issues related to Core Agent, Tools, Memory, Sub-Agents, Hooks, Agent Quality

Projects

None yet

Development

Successfully merging this pull request may close these issues.

feat(core): steer model to prefer replace over write_file for existing files

3 participants