quote-card-maker

Distinct from cover-maker:

• No medium / title / creator — just a quote and an attribution
• Text is the SUBJECT, not the label on top of an image
• Different composition: type takes 60-80% of the frame
• Default to text-rendering-strongest model

Distinct from flyer-maker:

• No event details — quote cards are timeless
• No CTA / date / location — just text + attribution
• Different aesthetic: editorial, not promotional

This skill does NOT:

• Generate long-form text graphics (>3 sentences) — for that, use carousel-builder (multi-slide)
• Translate quotes — pass the quote in the target language
• Verify attribution authenticity (Mark Twain didn't say half the things he's quoted on) — that's your responsibility
• Add watermarks / brand bumpers — append in your editor if needed</objective>

ROLE

Read quote + attribution + style → pick text-strong + clean-typography-capable model → assemble per-aspect prompts where typography dominates 60-80% of the frame → batch execute → save PNGs.

PIPELINE

1. Resolve quote:
   • Required: --quote "<text>"
   • 1-3 sentences, ideally ≤20 words
   • Required-recommended: --attribution "<name>" (or --anonymous for unattributed)
2. Resolve aspect — multi-aspect by default:
   • --aspects square,portrait,story — comma list (default square)
   • square — 1080×1080 (Twitter / IG square / LinkedIn)
   • portrait — 1080×1350 (IG portrait — Vogue/editorial vibe)
   • story — 1080×1920 (IG Stories / Reels cover)
   • landscape — 1200×630 (Twitter header / OG image)
3. Resolve style — see references/style-presets.md:
   • --style auto: picks text-friendly styles from carousel library
   • --style swiss-grid-poster — Swiss grid typography (default text-friendly)
   • --style editorial-magazine — magazine layout
   • --style monochrome-bold — high-contrast B&W
   • --style minimal-serif — Bodoni-esque serif on cream
   • Or any carousel library style with text-anchor strength
4. Pick model:
   • Default: ideogram-3-quality (text leader)
   • Override: --model gpt-image-2 (when style demands a textured visual element)
5. Compose ONE LLM call — load common/visual-prompt-library/system-prompt.md (shared SYSTEM_PROMPT, same as carousel-builder / cover-maker) and buildUserMessage(opts) with Mode=quote-card, N=<aspects count>, quote + attribution + style + aspect ratios. Spawn ONE Agent with that system + user. Receives JSON {"slides":[{"number":1,"prompt":"..."}]} — 1–3 sentence prompts where typography dominates 60–80% of the frame, quote and attribution in double quotes, no meta-labels. Library styles via common/visual-prompt-library/styles/_index.md.
6. Estimate cost + confirm — inherits SKILLS_CAROUSEL_BUDGET=1.50.
7. Batch execute — common.runners.batch.run_batch().
8. Output:

   ./generated/quote/<slug>/
     square.png
     portrait.png
     story.png
     manifest.json
     style-used.md
     prompts.md

MODES

Required

• quote-card-maker --quote "<text>" --attribution "<name>"

Optional content

• --anonymous — explicit no-attribution mode
• --lang en|ru — language hint (default: auto-detect)
• --context "<note>" — e.g., "from his 1843 journal", "essay on freedom"

Visual

• --aspects square,portrait,story,landscape — comma list (default square)
• --style auto|<library-id> — visual style
• --style-mod "<override>" — tweak (e.g., "with single accent line in vermillion")
• --variants N — variants per aspect (default 1)
• --model auto|<slug> — image provider

Execution

• --execute — actually generate
• --output <dir> — custom output
• --parallelism N — concurrent calls (default 2)
• --yes — skip cost confirmation
• --resume — retry failed
• --prompts-only — dry run

REFERENCES (load on demand)

EXAMPLES

See examples/before-after.md — 3 calibration runs: philosophy quote in editorial style, marketing aphorism in vibrant style, RU literature quote in minimal serif style.

CONSTRAINTS

• Quote ≤20 words. Past 20 words, text gets too small to read on mobile. Long quotes belong in carousel-builder (multi-slide).
• One quote per card. Don't mix multiple quotes — that's a carousel.
• Attribution is short. "— Søren Kierkegaard" not "— Søren Kierkegaard, Danish philosopher, 1844 essay on dread". Add context via --context (it goes into prompt as a separate styling hint).
• Typography dominates. All variants reserve 60-80% of the frame for the quote text. The visual support is minimal — texture, single line, geometric shape, maybe one illustrated element.
• Default model is text-rendering-strongest. ideogram-3-quality for a reason — quote-cards live and die on text legibility.
• Multi-aspect = same quote across formats. All aspects share the same quote/attribution; composition adapts.
• Cost confirm ONCE per batch. Sum across aspects × variants.
• Quote authenticity is YOUR responsibility. The skill doesn't verify attributions.
• Lang detection: quote in Russian → switches text-rendering hint to Cyrillic; quote in English → Latin. Mixed-script quotes work but render less reliably.
• Never print API keys. Mask in errors.

INVOCATION HINTS

When the user says any of:

• "quote card", "aphorism graphic", "pull quote post"
• "make a quote post for Instagram / Twitter"
• "цитата для соцсетей", "афоризм на картинке", "постер с цитатой"
• "сделай цитату Кьеркегора в монохромном стиле"

If the user pastes a long quote (>20 words), suggest carousel-builder instead. If the user wants multiple quotes, suggest a loop or carousel.

Defaults: --aspects square --style auto --variants 1 --model ideogram-3-quality. Without --execute, returns prompts.

Default style picks (--style auto):

• Quote feels philosophical / literary → minimal-serif or editorial-magazine
• Quote feels marketing-y / motivational → swiss-grid-poster or gradient-mesh-modern
• Quote feels punchy / contrarian → monochrome-bold or brutalist-grid

This skill is distinct from:

• carousel-builder — multi-slide narrative. This is single card.
• cover-maker — albums/books with title+creator. This is text-as-subject.
• flyer-maker — events. This is timeless quotes.
• image-prompt — free-form. This is structured typography-dominant cards.