[BUG] Drag-and-drop image file pastes path instead of attaching as [Image] (Warp Terminal, macOS)
Resolved 💬 3 comments Opened Apr 14, 2026 by bachar-ven Closed Apr 18, 2026
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?
Summary
Dragging and dropping an image file (e.g., a screenshot from Cmd+Shift+4) into the Claude Code CLI input pastes the raw file path as text instead of attaching the image inline as [Image #N].
This previously worked on a different machine (same terminal — Warp, same OS — macOS, likely older Claude Code version). Clipboard paste via Ctrl+V still works correctly.
Steps to Reproduce
- Take a screenshot with
Cmd+Shift+4(saves file to desktop/temp) - Open Claude Code CLI in Warp Terminal
- Drag and drop the screenshot file into the Claude Code input
Expected: Image is attached inline, displayed as [Image #1] chip.
Actual: The escaped file path is pasted as plain text:
/var/folders/.../Screenshot\ 2026-04-14\ at\ 4.35.10\ PM.png
Claude Code does not recognize it as an image.
Additional Testing
- Clipboard paste works:
Cmd+Ctrl+Shift+4(screenshot to clipboard) →Ctrl+Vin Claude Code → correctly shows[Image #1]. So thesharpimage processing pipeline is fine. - Manually typing the image file path in the input also does not trigger
[Image]detection — same behavior as drag-and-drop. - Both drag-and-drop and manual path entry worked on a previous machine (same Warp Terminal, macOS, likely older Claude Code version).
Environment
- Claude Code: 2.1.108
- macOS: 26.4 (Build 25E246)
- Architecture: arm64 (Apple Silicon)
- Terminal: Warp Terminal
- Shell: zsh
- Node.js: managed via mise
Possibly Related Issues
- #39416 — Multi-image drag-and-drop regression since v2.1.83 (only first image attached)
- #32950 — Drag-and-drop stopped working on Windows 11
🤖 Generated with Claude Code
This issue has 3 comments on GitHub. Read the full discussion on GitHub ↗