[BUG] allowed-tools in SKILL.md not parsed properly

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?

I created a SKILL.md file with the following header:

allowed-tools:
  - Read(~/vault/*)
  - Write(~/vault/*)
  - Edit(~/vault/*)
  - Bash(pgrep *)
  - Bash(obsidian open *)

Claude Code correctly invoked the pgrep tool without asking, but it still asked for permission to run obsidian open. The situation remained the same even after I changed it to Bash(obsidian *) only.

The actual commands invoked by Claude:

Bash(pgrep -x obsidian; echo "exit: $?")
Bash(obsidian open path="mydir/myfile.md" newtab)

What Should Happen?

Claude should run the allowed tool without asking.

Error Messages/Logs

(none)

Steps to Reproduce

1. Invoke the skill
2. Observe

Claude Model

Sonnet (default)

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.85 (Claude Code)

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

Other

Additional Information

_No response_