[BUG] kubectl command patterns not matching despite correct allow rules

Environment

• Platform (select one):
• [x] Anthropic API
• [ ] AWS Bedrock
• [ ] Google Vertex AI
• [ ] Other: <!-- specify -->
• Claude CLI version: 1.0.96
• Operating System: macOS (Darwin 24.6.0)
• Terminal: Bash in VSCode

## Bug Description
kubectl command patterns in .claude/settings.local.json allow rules are not matching despite correct wildcard syntax. Commands that should be automatically
allowed are still prompting for user permission.

## Steps to Reproduce

1. Add kubectl patterns to .claude/settings.local.json allow rules:

``json
"allow": [
"Bash(kubectl *get*)",
"Bash(kubectl *logs*)",
"Bash(kubectl *describe*)"
]
``

1. Execute simple kubectl commands like:

• kubectl get pods
• kubectl -n system-upgrade logs podname
• kubectl get nodes

1. Observe permission prompts appear

## Expected Behavior

Commands matching the wildcard patterns should execute automatically without permission prompts. For example:

• kubectl get pods should match Bash(kubectl get)
• kubectl logs podname should match Bash(kubectl logs)

## Actual Behavior

All kubectl commands prompt for user permission despite having matching allow patterns configured.

## Additional Context

• Even the simplest kubectl get pods command triggers permission prompt
• The patterns appear syntactically correct based on documentation
• Other bash command patterns in the allow list work correctly
• Workaround: Using broader pattern like Bash(kubectl*) might work but reduces granular control
• This affects kubectl debugging workflows significantly in Kubernetes environments