[BUG] Claude code only recognizes legacy `:*` suffix in `.claude/settings.json` file

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?

With "Bash(pnpm lint *)" as an entry in the "allow" permissions of .claude/settings.json, Claude Code is prompting to run the lint command. If I change this entry to "Bash(pnpm lint:*)", using the legacy :* suffix, it correctly runs the lint command without prompting.

Within a SKILL.md file, allowed-tools: Bash(pnpm lint *) works without prompting, so it seems like only skills are respecting the wildcard syntax without the legacy suffix.

What Should Happen?

According to the docs, the :* suffix is legacy and * should work in its place consistently now.

Error Messages/Logs

Steps to Reproduce

1. Create a .claude/settings.json file

{
  "$schema": "https://json.schemastore.org/claude-code-settings.json",
  "permissions": {
    "allow": [
      "Bash(pnpm lint *)"
    ]
  }
}

1. Create a new skill like .claude/skills/test/SKILL.md

---
allowed-tools: 
---

Run pnpm lint

1. Try triggering this skill and checking if the command is run without prompting
2. Change it to Bash(pnpm lint:*)
3. Try triggering this skill and checking if the command is run without prompting
4. Change allow in .claude/settings.json to be empty [] and in SKILL.md, add allowed-tools: Bash(pnpm lint *)
5. Try triggering this skill and checking if the command is run without prompting

Bash(pnpm lint *) works within SKILL.md, but not in settings.json.
Bash(pnpm lint:*) works in settings.json.

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.38

Platform

Anthropic API

Operating System

macOS

Terminal/Shell

Terminal.app (macOS)

Additional Information

_No response_