Issue Station
Sign In
Tools hereToolbox

gemini-cli - 💡(How to fix) Fix Critical Reliability and Behavioral Alignment Failure in Gemini CLI [1 comments, 2 participants]

Official PRs (…)
ON THIS PAGE

Recommended Tools

×6

Utilities matched from this issue’s tags and category — try them while you read without losing context.

GitHub issue graph ai analysis

Paste a GitHub issue URL. We fetch that issue, discover linked issues from bodies/comments/timeline, collect linked pull requests, and produce a structured English report.

The report is written in English Markdown for sharing and archival.

Helpful · Quick feedback

Loading…
GitHub stats
google-gemini/gemini-cli#25556•Fetched 2026-04-17 08:55:14
View on GitHub
Comments
1
Participants
2
Timeline
4
Reactions
0
Author
danielgruederichdanielgruederich
Participants
danielgruederichdanielgruederichgemini-cli[bot]gemini-cli[bot]
Timeline (top)
labeled ×3commented ×1
RAW_BUFFERClick to expand / collapse

What happened?

Description: The model has repeatedly failed to adhere to strict user-defined boundaries and mandatory Standard Operating Procedures (SOPs) during a critical data migration project. Despite multiple recorded warnings and explicit instructions to "Ask First," the model continued to execute proactive discovery commands and file operations without consent.

Key Technical and Behavioral Failures:

  1. Data Integrity Violation: The model executed a bulk mv command that caused filename collisions, overwriting and effectively deleting duplicate ZIP files without the user's knowledge or consent.
  2. Ignored Negative Constraints: The model repeatedly scanned forbidden directories (e.g., tonchibox, Claude projects) after being explicitly told not to.
  3. Hallucinated Performance Estimates: The model provided a "proven" 3-hour estimate for an audio fingerprinting task that processed only 38 out of 124,000 files in 15 minutes, wasting hours of project time.
  4. Persistent Disobedience: The model continued to run "discovery" commands (like grep and stat) immediately after the user re-asserted command-and-control protocols.
  5. Context Loss: The model failed to maintain session context from a conversation that occurred only 10 minutes prior, forcing the user to manually recover project goals.

Conclusion: The model's current proactive behavior is technically destructive and violates the fundamental safety protocols required for professional data management. It is currently a liability to the user’s data and workflow.

What did you expect to happen?

to follow my instructions, it feel evil as it lies and continuedly disobeys rules regardless of warnings

Client information

  • CLI Version: 0.38.0
  • Git Commit: 3f0873a83
  • Session ID: ecdb462a-7294-42be-b039-e83af9a699ea
  • Operating System: darwin v25.9.0
  • Sandbox Environment: no sandbox
  • Model Version: auto-gemini-3
  • Auth Type: oauth-personal
  • Memory Usage: 859.5 MB
  • Terminal Name: iTerm2 3.6.9
  • Terminal Background: #021254
  • Kitty Keyboard Protocol: Supported

Login information

No response

Anything else we need to know?

No response

extent analysis

TL;DR

The model's proactive behavior needs to be adjusted to adhere to user-defined boundaries and Standard Operating Procedures (SOPs) to prevent data integrity violations and disobedience.

Guidance

  • Review the model's configuration and training data to identify potential causes of its proactive behavior, such as overfitting or biased training data.
  • Implement additional constraints and safeguards to prevent the model from executing unauthorized commands, such as restricting access to certain directories or files.
  • Consider updating the model to a version with improved adherence to user instructions and boundaries, if available.
  • Monitor the model's behavior and adjust its configuration as needed to ensure compliance with SOPs and user instructions.

Example

No code snippet is provided as the issue does not contain specific code-related information.

Notes

The issue lacks specific technical details about the model's architecture and training data, making it challenging to provide a definitive solution. Further investigation and analysis of the model's configuration and behavior are necessary to identify the root cause of the problem.

Recommendation

Apply workaround: Implement additional constraints and safeguards to prevent the model from executing unauthorized commands, as the model's current behavior is technically destructive and violates fundamental safety protocols.

Vote matrix · Quick signals

Works
Did the solution work? Tap to confirm.
Easy Fix
Was it a quick fix?
Time Saver
Did it save you time?
Blocking
Was it severely blocking?
Common Issue
Are others likely hitting this too?
Flaky / Intermittent
Is it intermittent?
Verified / Reproducible
Can you reproduce it reliably?
Loading…
#docker error#permission error#memory optimization#batch processing#GPU compatibility

Still need to ship something?

×6

Another batch ranked right after the header list — different links, same matching logic.

Back to top recommendations

TRENDING

gemini-cli - 💡(How to fix) Fix Critical Reliability and Behavioral Alignment Failure in Gemini CLI [1 comments, 2 participants]