[FEATURE] Render images inline via sixel/kitty graphics protocol in terminal (CLI)

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

Feature request

When Claude Code (CLI) reads an image file (chart, screenshot, plot), the
image is visible to the model but not to the user sitting in the terminal.
Over SSH to a headless box there is no desktop app or browser to fall back
to, so the user ends up scp-ing files around or blindly trusting Claude's
verbal description of the image ("the loss curve looks fine, trust me").

Proposed Solution

when the terminal advertises sixel or kitty graphics support
(via terminfo/XTGETTCAP or an opt-in setting), render images that Claude
reads (or a new "show image" action) inline as DCS sequences in the
transcript.

Alternative Solutions

_No response_

Priority

Low - Nice to have

Feature Category

CLI commands and flags

Use Case Example

I run long darknet/YOLO training jobs on a remote
GPU box and use Claude Code over SSH to babysit them. Claude can already
Read the loss chart PNG and analyze the curve shape — but I can't see it.
Windows Terminal, wezterm, foot, mlterm etc. all support sixel today, so
inline display would close the loop without any GUI or file transfer.

Note: piping img2sixel through the Bash tool doesn't work as a
workaround — tool output is captured and escaped, so the DCS sequences
never reach the terminal raw. This needs renderer support.

Additional Context

_No response_