Support skill metadata to invoke subagents

### Describe the feature or problem you'd like to solve

An equivalent to claude's `context: fork` metadata in skills

### Proposed solution

https://code.claude.com/docs/en/skills#run-skills-in-a-subagent

I have a main agent file that acts as a workflow that delegates to subagents by explicitly stating in the agent file `"Step 1: Run subagent to follow instructions in prompt file at c:/path"`

My wish is to turn each step into a skill instead of a prompt, and it would be great for those skills to define "this should be run in a subagent", because those skills are effectively instructions to follow

### Example prompts or workflows

_No response_

### Additional context

_No response_

先閱讀連結的 Claude 文件中關於在 subagent 中執行的 skill 的內容，然後檢查此 CLI 中現有的 skill 和 subagent 呼叫行為。完成標準是：skill 可以宣告要求在 subagent 中執行的中繼資料，並且 workflow 會據此呼叫它。

由索引模型根據 Issue 內容生成。