give AI context about your images.

imgstat scans your project and generates image dimension reports your AI coding assistant can actually read — without guessing, without vision tokens, without touching your source files.

npm install -g imgstat

imgstat infers your intent automatically based on the target you provide.

# Print dimensions of images from a web URL
imgstat https://example.com/logo.png

# Inspect a local directory
imgstat ./public/images

# Output strict, flat JSON (ideal for AI parsing)
imgstat https://example.com/images --json

# Rename local images to append dimensions (e.g., image-800x600.jpg)
imgstat rename ./public/images -y

When run in non-interactive environments or with flags, imgstat routes all progress logs to standard error, ensuring stdout remains clean for pipeline parsing and LLM usage.

Keep the tool small. If you are considering adding a feature, ask: does this help AI understand images faster? If the answer is not clearly yes, it probably does not belong here.

Every file in lib/ must have one clear responsibility. If you find yourself writing image discovery logic inside rename.sh, stop and move it to scan.sh.

No feature should require memorizing new flags. If it can be handled by a mode or an interactive prompt, prefer that.

Remote mode must never leave files on disk. The trap cleanup is non-negotiable.

Dry-run must work for any operation that touches files. This is a safety contract with users.