Issue Station
Sign In
Tools hereToolbox

claude-code - 💡(How to fix) Fix Claude Code web session: agent implemented plan instead of returning to terminal, burning usage on unretrievable code [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
anthropics/claude-code#56046•Fetched 2026-05-05 05:59:36
View on GitHub
Comments
1
Participants
2
Timeline
6
Reactions
0
Author
tblachektblachek
Participants
github-actions[bot]github-actions[bot]tblachektblachek
Timeline (top)
labeled ×5commented ×1

Code Example

None. It generated dead unretrievable code in the cloud while I was working through the implementation myself within the terminal locally.

---

i didn't tell you to code in this session, i approved the plan to be sent back to the terminal.

You're right, and I apologize. The plan was approved to be implemented back in your terminal, not here. I should not have started coding — I burned significant usage generating code in an environment where it can't be committed or retrieved.

I don't have the ability to issue refunds or credits. For that, you'd need to contact Anthropic support or report the issue at https://github.com/anthropics/claude-code/issues.
RAW_BUFFERClick to expand / collapse

Preflight Checklist

  • I have searched existing issues for similar behavior reports
  • This report does NOT contain sensitive information (API keys, passwords, etc.)

Type of Behavior Issue

Claude modified files I didn't ask it to modify

What You Asked Claude to Do

Terminal session produced a Spec, Spec was brought to browser with /ultraplan. Plan was generated and revisions were made. The plan was ultimately 'accepted' and sent back to the terminal.

What Claude Actually Did

While I moved back to terminal, the browser agent decided to try implementing the plan blindly in the background.

Session ID: session_017dCufBj6FwWDV1hgZkLjmn

  • The agent resumed from a context compaction and began implementing the full plan (32 files, ~2400 lines) in the web session instead
  • The commit signing server was down, so nothing could be committed
  • No git remote was configured, so code can't be pushed
  • The generated code is effectively unretrievable
  • Request: usage credit/refund for the implementation work that should not have been performed

Expected Behavior

I expected the /ultraplan session to send the plan back to the terminal for me to begin implementing within my workflow and end the session.

Files Affected

None. It generated dead unretrievable code in the cloud while I was working through the implementation myself within the terminal locally.

Permission Mode

Accept Edits was OFF (manual approval required)

Can You Reproduce This?

Haven't tried to reproduce

Steps to Reproduce

No response

Claude Model

Opus

Relevant Conversation

i didn't tell you to code in this session, i approved the plan to be sent back to the terminal.

You're right, and I apologize. The plan was approved to be implemented back in your terminal, not here. I should not have started coding — I burned significant usage generating code in an environment where it can't be committed or retrieved.

I don't have the ability to issue refunds or credits. For that, you'd need to contact Anthropic support or report the issue at https://github.com/anthropics/claude-code/issues.

Impact

Low - Minor inconvenience

Claude Code Version

Claude Code version 2.1.126.

Platform

Anthropic API

Additional Context

No response

extent analysis

TL;DR

The issue can be mitigated by ensuring that the /ultraplan session is properly terminated or paused when sending the plan back to the terminal for implementation.

Guidance

  • Review the documentation for the /ultraplan command to understand how to properly terminate or pause the session.
  • Verify that the "Accept Edits" mode is set to "OFF" (manual approval required) to prevent unintended changes.
  • Check the Anthropic API documentation for any specific guidelines on handling plan implementation and session management.
  • Consider contacting Anthropic support for assistance with issuing a refund or credit for the unintended usage.

Notes

The root cause of the issue appears to be a misunderstanding of the /ultraplan session behavior, and the fix may involve adjusting the workflow to properly manage the session.

Recommendation

Apply workaround: Ensure that the /ultraplan session is properly terminated or paused when sending the plan back to the terminal for implementation, as this will prevent unintended changes and usage.

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…
#api#tool integration#LLM response#prompt template#agent execution

Still need to ship something?

×6

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

Back to top recommendations

TRENDING

claude-code - 💡(How to fix) Fix Claude Code web session: agent implemented plan instead of returning to terminal, burning usage on unretrievable code [1 comments, 2 participants]