[BUG] Permission patterns in settings.json not being respected - commands still prompt for permission

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?

Environment:

Claude Code version: (2.1.15 (Claude Code))
OS: Linux (RHEL 9.5)
Shell: bash
Description:
Commands that match explicit allow patterns in ~/.claude/settings.json still trigger permission prompts.

Steps to reproduce:

Create ~/.claude/settings.json with:

{
"permissions": {
"allow": [
"Bash(ls*)"
]
}
}
Ask Claude to run ls -la /some/path

Claude prompts for permission instead of executing directly

Expected behavior:
Command ls -la /some/path should match pattern Bash(ls*) and execute without prompting.

Actual behavior:
Claude prompts for permission despite the command matching the allow pattern.

Additional context:

The same issue occurs with other patterns like Bash(ctest*)
Claude acknowledges it has permission ("I have Bash(ls*) permission") but then immediately asks for permission anyway
This happens consistently after context compaction/summarization, but also in fresh sessions
Workaround attempted:
Changed patterns from "Bash(ctest )" (with space) to "Bash(ctest)" (no space) - did not resolve the issue.

What Should Happen?

see above (composed by claude itself)

Error Messages/Logs

see above (composed by claude itself)

Steps to Reproduce

see above (composed by claude itself) , may be you should give claude ability to report bugs , instead of making paying (!!!) customer jump through these hoops

Claude Model

Opus

Is this a regression?

Yes, this worked in a previous version

Last Working Version

na

Claude Code Version

2.1.15 (Claude Code)

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

it is getting increasingly frustrating when claude asks permissions every 10 seconds for something it has permission to do , it cannot explain itself why it does that . Add to this the fact that it compacts all the time and after compacting looses context and permissions - I am not sure , but I think it is very strong point of switching to any other ai app