The image is at artifacts/image.png: a 512 × 512 PNG showing a brick-red bear bust in black-framed shades and a blue robe on the dark green-black background.

How I made it: I drew the bear pixel by pixel on a 32 × 32 grid in a Python script, tools/generate_iu_0327.py, then scaled it up 16 times with hard pixel edges. I wrote it this way rather than using the image-generation tool so the exact colours and grid are guaranteed. The script needs nothing outside standard Python, so it runs without a network.

What's in the picture:

• One front-facing bear, centred, with the robe's shoulders reaching the bottom edge.
• Brick fur (#df523e), rust inner ears and a tangerine muzzle with a black nose and mouth.
• Shades with navy lenses and small bone-white glints. No headgear.
• A cobalt robe with a gold V-neck trim and navy fold lines.
• A plain #0c1610 background with five single-pixel green specks. No text, frame or border.

Checks: The image-inspection tool (ffprobe) wasn't installed, so I wrote a small script that decodes the PNG directly. It confirmed:

• 512 × 512, standard 8-bit colour PNG.
• Every 16 × 16 block is a single flat colour, so there is no anti-aliasing, gradient or blur.
• Exactly 10 colours: 9 from your palette plus black for the outline. Using black is the one departure from the palette list, since the brief also asks for a black outline.

Not done:

• The outline is one grid pixel thick, which is 16 pixels in the final image.
• Nobody has checked how well it matches the rest of the 787-piece set. I only looked at it myself.

I committed the script and a README.md recording all of this. The image itself is left out of git as required: the workspace's local git settings already ignore artifacts/.