[FEATURE] Separate flags for showing thinking vs tool results in verbose mode

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Currently --verbose shows both extended thinking AND full tool results together. There's no way to see Claude's thinking process without being flooded with tool output (file reads, grep results, etc.).

Related issues #8477 and #8371 request showing thinking by default, but this is a different request: granular control over what verbose mode displays.

When debugging Claude's reasoning or understanding why it made certain decisions, I want to see the thinking process. However, tool results (especially large file reads and grep outputs) create significant noise that makes it hard to follow the actual reasoning.

Proposed Solution

Option A: Separate flags

# Show only thinking
claude --show-thinking

# Show only tool results  
claude --show-tool-results

# Show both (current --verbose behavior)
claude --show-thinking --show-tool-results

Option B: Verbose with value (preferred)

claude --verbose=thinking    # Only thinking blocks
claude --verbose=tools       # Only tool results
claude --verbose=all         # Both (current behavior)
claude --verbose             # Defaults to 'all' for backwards compatibility

Option B is cleaner as it extends the existing flag rather than adding new ones.

Alternative Solutions

Current workaround is piping through grep to filter output:

claude --verbose 2>&1 | grep -v "Tool result"

This is fragile and doesn't work well with the TUI.

Priority

Medium - Would be very helpful

Feature Category

CLI commands and flags

Use Case Example

1. I ask Claude to refactor a complex function
2. Claude reads multiple files, searches the codebase, then makes changes
3. I want to understand WHY Claude chose a particular approach
4. With current --verbose, I see 500+ lines of file contents mixed with thinking
5. With --verbose=thinking, I'd see only the reasoning process - much easier to follow

Additional Context

This would complement the existing requests for better thinking visibility (#8477, #8371) by providing more fine-grained control for users who want thinking but find full tool output overwhelming.

---

Submitted by @halilc4
This issue was drafted with assistance from Claude Code (Opus 4.5)