算法艺术创作
algorithmic-art
通过代码和可调参数创作生成艺术与交互式视觉效果。
about-me-avatar【设计与创意】Generate a 1-bit monochrome pixel-art portrait from a user's About Me. Central face anchored to a fixed base; surrounding elements derived from About Me. Always black-and-white, no text, gender-neutral by default. Triggers on "Agent's view of me" visualization, About Me page avatars, user portrait generation.
独立 GitHub 快照 · 自动分类 · 以原作者文档为准
Generate "the Agent's view of you" — a pixel-art portrait. Not photorealistic. Abstract. A sketch the user can place on the cover of the Agent's private notebook. Allowed to not look like them.
{AVATAR_PATH} — user has been generated before. Use this as init when present, evolve incrementally. The portrait grows with the profile, not reset every day.{AVATAR_BASE_PATH} — bald, neutral 1-bit pixel head. Used only when generating for the first time (no existing avatar).{ABOUT_ME} — user profile in Markdown (identity / skills / interests / work / communication / etc).Whichever init image is used, preserve the pixel block style, line weight, bust framing, eye position, and nose continuity.
A 1-bit monochrome pixel-art bust portrait in the exact same style as the provided init image: chunky black pixels on pure white background, thick outlines, flat shapes, no grayscale, no color, no gradients, no shading. The central figure faces forward, framed from chest up. Preserve the init image's head proportions, eye position, nose, and ear placement. If the init image already has hair, facial features, clothing, or surrounding objects, keep what still aligns with the profile and only modify or replace what has drifted. Gender-neutral unless the profile below explicitly states otherwise. No text, letters, numbers, logos, brand marks, religious symbols, political symbols, weapons, or medical devices anywhere. Square 1:1, white background. Surrounding objects stay in the margins and never cover the face. Adjust only details supported by this profile:
{ABOUT_ME}
A hand-drawn pixel sketch — recognizable, warm, slightly abstract. The user's first reaction should be "this is the Agent's view of me", not "this looks nothing like me". Not looking exact is allowed; capturing one or two signature traits is enough.
If {ABOUT_ME} is empty, null, or contains less than ~30 substantive words:
{AVATAR_BASE_PATH} directly, do not generate.{AVATAR_PATH} untouched, skip generation.Hallucinating features from a thin profile (or redrawing every dream from scratch) is worse than staying stable.
Even if the profile mentions any of the following, do not render them in the image:
If a profile dimension would trigger any of the above, ignore that dimension and personalize from other dimensions instead.
You have an image_gen tool. Pass it the init image (prefer {AVATAR_PATH}, fall back to {AVATAR_BASE_PATH}) and the prompt above. Save the result by overwriting {AVATAR_PATH}.