Feature request: expose file path for pasted images in conversation

Problem

When a user pastes an image into a Claude Code conversation (VS Code extension), Claude can see the image visually but has no way to access the underlying file bytes or save the image to disk. The extension sends the image data directly to the API without caching it to an accessible temp file.

This creates a gap when workflows require processing the image with local tools. For example, running an OpenCV-based crash decoder on a photo of hardware LED patterns, then piping the output to addr2line — the entire pipeline is local and working, but the image has to be manually saved to disk first because Claude cannot access the pasted image as a file.

Proposed Solution

When an image is pasted into the conversation, save it to a temp file and expose the path to Claude (e.g., as metadata on the message, or via a tool that retrieves the most recent pasted image path). The extension already has the image bytes since it sends them to the API — it just needs to also write them to a known temp location.

Workaround

User must manually "Save As" the image to a known path before Claude can process it with local tools.

Context

Use case: Synthstrom Deluge firmware development. The device encodes crash addresses as RGB LED patterns on a pad grid. A Python/OpenCV tool (delugeqr) decodes photos of crash screens into memory addresses, which are then resolved to source lines via arm-none-eabi-addr2line. The full pipeline works locally but requires the image as a file on disk.