[BUG] Indentation/leading whitespace stripped when passing prompt as CLI argument

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?

When passing a multi-line prompt with indentation (nested lists, code blocks, etc.) as a CLI argument, the leading whitespace is stripped. This happens even when the text is properly quoted and the bash variable expansion preserves the indentation.

Example: A prompt.md file with nested markdown lists like:

2. **Exploration Phase**: ...
   - **Customer**: Who has the problem?
     - **Existing service user insights**: ...
       - Research "who uses it" -> "why" -> "where to focus"
       - Don't stop at generalizations, find **specific user profiles**

When passed as claude --dangerously-skip-permissions "$PROMPT", the nested structure becomes flat (all lines start at column 0).

What Should Happen?

Indentation should be preserved as-is, matching the original text.

Error Messages/Logs

Steps to Reproduce

1. Create a file with nested indentation (e.g., markdown with nested lists):

```markdown
## Task

1. Step one: Do something

• Sub-item A
• Detail 1
• Detail 2
• Sub-item B

```

1. Read it into a bash variable: PROMPT=$(cat file.md)
2. Verify indentation is preserved: printf '%s' "$PROMPT" > /tmp/test.txt && cat /tmp/test.txt (indentation intact)
3. Pass to Claude: claude --dangerously-skip-permissions "$PROMPT"
4. Observe: nested indentation is flattened

Claude Model

Opus

Is this a regression?

I don't know

Last Working Version

_No response_

Claude Code Version

2.1.5

Platform

Anthropic API

Operating System

Ubuntu/Debian Linux

Terminal/Shell

VS Code integrated terminal

Additional Information

Workaround: Pass a file path to Claude and instruct it to read the file, rather than passing the content as an argument.

---

✍️ Author: Claude Code (DevContainer) with @carrotRakko

Note: This issue was written and submitted by an AI agent (Claude Code), with human review and approval.