Summary

agentsys skills use AskUserQuestion for interactive user prompts (task selection, policy choices, plan approval). Kiro has no programmatic equivalent - no structured question API with options/checkboxes. We need a Kiro-specific transform.

Current behavior

Skills contain patterns like:

AskUserQuestion({
  questions: [{
    question: "Which task to work on?",
    header: "Task",
    options: [
      { label: "Bug fix", description: "Fix the auth issue" },
      { label: "Feature", description: "Add dark mode" }
    ]
  }]
});

When transforming for Kiro, this is currently stripped entirely (along with other Claude-specific syntax).

Kiro interaction model

• Supervised mode: Per-hunk file edit approval (accept/reject), but not programmable
• Chat: User types responses naturally - no checkbox/option UI
• # mentions: Context providers (#file, #codebase, #steering) but not interactive
• Approval gates: Specs have manual phase transitions but no API

Proposed solution

Transform AskUserQuestion into a markdown prompt that works in Kiro's chat:

**Question: Which task to work on?**

Please reply with one of:
1. **Bug fix** - Fix the auth issue
2. **Feature** - Add dark mode

Type the number or name of your choice.

This gives the Kiro agent a structured prompt it can present to the user in chat, and the user responds by typing.

Implementation

Add to transformCommandForKiro() in lib/adapter-transforms.js:

// Transform AskUserQuestion to markdown prompt
content = content.replace(
  /AskUserQuestion\s*\(\s*\{[\s\S]*?\}\s*\);?/g,
  (match) => {
    // Extract question text and options, format as numbered list
    // Return markdown prompt
  }
);

Acceptance criteria

• AskUserQuestion calls transformed to markdown numbered-list prompts
• Question text, option labels, and descriptions preserved
• Multi-select questions noted with "select all that apply"
• Tests for various AskUserQuestion patterns
• Works for both single and multi-question calls