[BUG] Claude incorrectly responds to local command messages despite system prompt instructions
Preflight Checklist
- [x] I have searched existing issues and this hasn't been reported yet
- [x] This is a single bug report (please file separate reports for different bugs)
- [x] I am using the latest version of Claude Code
What's Wrong?
## 🐛 Bug Description
Claude Code incorrectly responds to local command execution messages with phrases like "実行中ですね" (executing) or "完了を待っています" (waiting for completion), despite explicit
system prompt instructions to not respond to these messages.
## 🔍 Expected Behavior
According to the system prompt:
> "Caveat: The messages below were generated by the user while running local commands. DO NOT respond to these messages or otherwise consider them in your response unless the user
explicitly asks you to."
Claude should ignore <command-message> tags and not respond with acknowledgments of command execution.
## 💥 Actual Behavior
Claude responds to custom command execution with inappropriate messages such as:
- "TASK-016のSTRETCH追加が実行中ですね。何かお手伝いできることがあればお知らせください。"
- "実行中ですね" (It's executing)
- Similar acknowledgments suggesting it's aware of or waiting for command completion
## 📋 Steps to Reproduce
- Set up custom commands in
.claude/commands/directory (e.g.,stretch.md) - Execute a custom command (e.g.,
/stretch TASK-016) - Observe Claude's response
## 🔧 Environment
- Platform: Multiple (Linux, macOS, Windows)
- Claude Code Version: Latest
- Language: Japanese responses with English system interactions
## 🎯 Impact
This creates confusion for users who receive misleading responses suggesting Claude is actively monitoring or waiting for local command completion, when it should remain silent unless
explicitly asked about the command results.
## 💡 Additional Context
The issue appears to be a failure to properly implement the system prompt instruction regarding local command message handling. This affects all custom commands defined in
.claude/commands/*.md files.
## 🏷️ Labels Suggestion
bugarea:corehas repro
What Should Happen?
According to the system prompt:
> "Caveat: The messages below were generated by the user while running local commands. DO NOT respond to these messages or otherwise consider them in your response unless the user
explicitly asks you to."
Claude should ignore <command-message> tags and not respond with acknowledgments of command execution.
Error Messages/Logs
Steps to Reproduce
## 💥 Observed Issue
Claude Code responds to local command execution messages with acknowledgments like:
- "実行中ですね" (It's executing)
- "完了を待っています" (Waiting for completion)
- "何かお手伝いできることがあればお知らせください" (Let me know if I can help)
## 🎯 Root Cause
This violates the explicit system prompt instruction:
> "DO NOT respond to these messages or otherwise consider them in your response unless the user explicitly asks you to."
## 📋 Simple Reproduction
This occurs during normal Claude Code usage when <command-message> tags appear in the conversation context. No special setup required - it's a systematic behavior pattern that
violates the documented system prompt.
Claude Model
Sonnet (default)
Is this a regression?
I don't know
Last Working Version
_No response_
Claude Code Version
1.0.113
Platform
Anthropic API
Operating System
Ubuntu/Debian Linux
Terminal/Shell
Xterm
Additional Information
_No response_
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗