Input Image Array (ShellAgent Plugin)

Input Image Array (ShellAgent Plugin) is the batch-capable cousin of Input Image: instead of one image, it accepts an array of images - a JSON list of URLs or file paths, the format a ShellAgent app uses when it hands a workflow several pictures at once - and splits it into the four things downstream nodes actually want: a batched IMAGE, batched masks, an iterable list of images, and a count.

This is the node you reach for when the app's input isn't "here's a picture" but "here's a pile of pictures" - a folder of references for style transfer, multiple input frames for a video model, a set of images to process in a loop.

How it works

The default_value is a JSON string, defaulting to [], with a placeholder showing the expected shape: ["https://example.com/a.png", "https://example.com/b.png"]. Each entry is loaded (local paths and URLs both work), and the results are organized into:

• images_batch - one IMAGE tensor with all frames stacked along the batch dimension.
• masks_batch - the matching MASK tensor for each image.
• images_list - the same images as a Python list (OUTPUT_IS_LIST true), for nodes that iterate.
• count - an INT with the number of images.

The resize_mode option decides what to do when the images aren't all the same size: pad_to_first, resize_to_first, or none_keep_list_only (leave them alone, useful only for the list output). min_items/max_items let you enforce the array length - handy as a sanity check against the app sending nothing or sending too much.

There's a deliberate edge-case behavior worth knowing: an empty array doesn't error - it returns a 1×1 black image and count 0, so the rest of the workflow doesn't crash while the app is still deciding what to send.

Inputs that matter

• input_name - the variable name the app sees. Default input_images.
• default_value - the JSON array of image references. This is what the app overrides at runtime.
• resize_mode - how to reconcile mixed sizes. resize_to_first (the default) is usually what you want for batching.
• min_items / max_items - optional length constraints; 0 means unlimited.

Installing it

Same one-pack install:

cd ComfyUI/custom_nodes
git clone https://github.com/myshell-ai/ComfyUI-ShellAgent-Plugin

restart, or grab ComfyUI-ShellAgent-Plugin from ComfyUI Manager. No model downloads; the pack's usual pip requirements install with it.

Common issues

• "Invalid array input" error. The default_value isn't parseable JSON. Watch for trailing commas or single quotes - JSON, not Python.
• Batch is all black. An empty (or all-failed) array falls back to the 1×1 black placeholder. Check that the app is actually sending entries.
• Images with different sizes break the batch. Use resize_to_first (or pad_to_first if you want to preserve aspect content) instead of the list-only mode.
• count is 0 but I sent images. The node truncates to max_items and validates against min_items; verify those aren't fighting your input length.

If you're building anything batch-shaped for an agent to feed, this is the input node you'll use far more than the singular Input Image.