Aesthetics Wiki

Aesthetics Wiki

A local, self-contained, queryable corpus of internet visual aesthetics. Each aesthetic is one markdown file with structured frontmatter and a full description. Everything needed is in this folder; nothing is fetched from the web. Do not load the whole corpus. Use the lookup CLI to find the right aesthetic, then read only that one file.

Data layout

• data/index.json - one compact record per aesthetic (name, aka, decade, colours, palette, motifs, values, related, subgenres, platforms, summary). Large; query it with the CLI rather than reading it whole.
• aesthetics/<slug>.md - full entry. Frontmatter fields: name, aka, decade_of_origin, creators, key_motifs, key_colours, palette, key_values, related_aesthetics, subgenres, primary_platform, related_media, source_url. Body is the full description (history, fashion, media, etc.). palette is a list of hex colours derived from the aesthetic's reference imagery (ranked by coverage) - use it as the real, buildable colour basis; key_colours is the prose description of the palette.
• data/palettes.json - the same palettes with coverage percentages.

Lookup CLI (use this first)

Run from the skill root. Prefer it over grepping raw JSON.

python scripts/lookup.py "dark academia"     # smart: resolves a name, else searches
python scripts/lookup.py --get cottagecore   # one entry, verbose, with its file path
python scripts/lookup.py --find "green forest witch"   # full-text across all fields
python scripts/lookup.py --color pastel      # aesthetics whose palette matches
python scripts/lookup.py --decade 1990s      # aesthetics of an era
python scripts/lookup.py --motif mushroom    # aesthetics sharing a motif
python scripts/lookup.py --related Cottagecore   # its related aesthetics + subgenres
python scripts/lookup.py --list              # every slug + name
python scripts/lookup.py --stats             # coverage counts

Add --json for structured output, --limit N to widen results.

After the CLI points you to a file, read aesthetics/<slug>.md for the depth (history, fashion, music, decor, values) you need to answer well.

How to handle common requests

• "What aesthetic is this / what's it called when..." - pull the distinctive cues (colours, era, objects, mood) from the user, run --find with them, and present the top 2-3 candidates with a one-line contrast. Confirm with the user before committing to one.
• "Describe / explain " - --get <name>, then read the file and summarise its origin, key motifs, colours, values, and related aesthetics.
• "Show me aesthetics like X" / adjacent vibes - --related X, and also --find on X's colours/motifs to surface cousins that are not cross-linked.
• Moodboard / style guide - read the entry and list palette, key_motifs, key_values, and related_media. For a shareable image, run python scripts/make_moodboard.py <slug> (palette + motifs, no API, no third-party photos). If the user has configured an image API in .env (see README), run python scripts/gen_moodboard.py <slug> for a photo-based moodboard instead; fall back to make_moodboard.py when no key is set.
• "Style my <outfit/room/brand/playlist/website> as " - translate key_motifs, key_values, and the hex palette into concrete choices for that medium. For UI/web/brand work, use palette directly as design tokens (CSS variables, a Tailwind theme, etc.), pairing motifs/values into type, layout, and imagery direction. Cite the aesthetic and offer 1-2 related aesthetics to blend.
• Combine two aesthetics - read both entries, find shared and contrasting motifs/colours, and propose a coherent fusion; name the overlap if a subgenre already covers it.
• An aesthetic that is not in the index - say it is not in this knowledge base rather than guessing, then offer the closest matches from --find.

Attribution (required when you reproduce content)

The descriptive text is adapted from the Aesthetics Wiki, licensed CC-BY-SA 4.0. When you quote or closely paraphrase an entry in a user-facing deliverable, credit the Aesthetics Wiki and link the entry's source_url. Your own synthesis, recommendations, summaries, and generated moodboards do not need attribution.