happycapy-ai/happycapy-skills/neon-motion

Neon Motion PPTX Skill

A reusable fixed-style PowerPoint generation skill. It builds on the pptx skill's technical capabilities while locking in the Neon Motion visual identity: a premium cinematic pitch built from editorial motion photography and restrained typography — full-bleed motion imagery in electric cyan versus warm orange light, thin geometric sans titles, hairline rules, and generous empty space.

Core principle — restraint and cinema win over content volume. When the user's topic carries more information than the format can hold gracefully, compress it into sparse typographic rows, one compact chart, a timeline, concise metric moments, and extra slides. The deck must feel like a premium cinematic pitch for the user's topic — not a dark SaaS dashboard, an AI report, a cyberpunk poster, a conference deck, a generic startup template, a cluttered product report, or a text-heavy memo.

This skill is topic-agnostic. It generates Neon Motion decks for any user-provided topic. Never hardcode a sample topic (a company name, AI customer intelligence, sales pitches, revenue operations, a specific year, fictional metrics, fixed pricing, or fixed product features) into a deck. Any topic-specific wording in this file is an example only — replace it at runtime with the user's topic variables.

Step 1: Gather Inputs (Ask First — Max 3 Questions)

Do not start generating immediately. Use the AskUserQuestion tool to present the missing questions in a single call (up to 3 at once). Only ask what the user hasn't already provided — if they gave the topic, slide count, or image choice, skip that question. Don't ask extra questions unless the deck literally cannot be built without them.

Question 1 — Topic (skip if already provided):

• header: "Topic"
• question: "What is the topic, product, company, or title of the deck?"
• options: 2–3 relevant example topics plus allow Other

Question 2 — Slide count (skip if already specified):

• header: "Slides"
• question: "How many slides should the deck include?"
• options: 8 slides, 10 slides (Recommended), 12 slides, 15 slides

Question 3 — Images:

• header: "Images"
• question: "How should visuals be handled?"
• options:
• Generate with gpt-image-2 — Cinematic cyan/orange motion photography + topic theme visuals (Recommended)
• Use my uploaded / reference images — Insert images I provide
• No generated images — Native typography, hairlines, and charts only (color-field plates stand in for photos)

If the user hasn't provided data, dates, audience, or other details, infer neutral content or use clearly labeled placeholder figures (see Content Rules). Do not keep asking. Once you have the three core answers, proceed to generation.

Note on the no-images choice: the Neon Motion style leans heavily on cinematic photography. If the user picks No generated images, replace every full-bleed photo with a full-bleed cyan→orange color-field plate (layered translucent CYAN/ORANGE/PETROL shapes) so the motion-lit feel survives — never leave photo zones blank or fall back to a plain dark canvas.

Step 2: Topic Translation — Everything Becomes a Cinematic Pitch

Before writing any slide copy, translate the user's topic into a Neon Motion frame. Every subject becomes a restrained, image-led pitch sequence — while preserving the user's actual subject.

Decide these before writing slides:

• The motion energy for the imagery (what kind of anonymous human/topic motion moves through cyan-and-orange light).
• The one-line cover title (thin white sans, no tagline).
• The wordmark (a product/company name, if any — placed tiny, never merged into the big title unless the user asks).

Step 3: Generate Images (only if the user chose gpt-image-2)

If the user chose No generated images, skip generation and use color-field plates (Step 1 note). For uploaded images, place them full-bleed or in small crops/mockup frames, and use color-field plates where an image is missing.

If the user chose gpt-image-2, read the generate-image skill at /home/node/.claude/skills/generate-image/SKILL.md for the exact API call pattern. Use openai/gpt-image-2 with response_format: "b64_json" and aspectRatio matching placement ("16:9" full-bleed, "4:3" thumbnails/mockups). Save each image to /tmp/neon-motion-img-N.png.

Generate two clearly separated asset types (these labels are production notes only — never printed on a slide). Both must contain no readable text, no logos, no brand marks, no readable faces.

Decorative image set — editorial long-exposure / cinematic motion photographs for the cover, full-bleed section plates, quote plates, and closing:

Cinematic editorial long-exposure photograph, anonymous people (or [TOPIC-RELEVANT SUBJECTS]) moving laterally
through electric cyan and warm orange light, blurred silhouettes and motion energy, natural cinematic lighting,
soft reflections, strong spatial depth, bright and vivid. No text, no logos, no readable faces, no dashboards,
no product UI, no lasers, no glass shards, no cosmic scenes, no tunnels, no neon grids, no empty interiors. 16:9.

Theme image set — restrained topic-support visuals for content slides (fictional interface mockups, minimal product-support scenes, anonymous team/workspace photography, other topic-specific support):

Restrained topic-support visual for [TOPIC], cinematic cyan-and-orange palette, calm and premium, anonymous
subjects only, natural lighting. No readable UI text, no logos, no brand marks, no readable faces, no legible
numbers, no charts. 4:3 or 16:9, editorial, non-literal.

Theme assets support the story; they never replace native editable charts, text, or structured content. Never use gpt-image-2 for full-slide screenshots, slide mockups with embedded text, text-heavy graphics, charts with embedded numbers, logos, or brand marks — those are always native PowerPoint.

Count guidance: 8 slides → 3–4 images · 10–12 slides → 4–6 · 15 slides → 6–8. Keep most of them decorative motion plates.

Step 4: Build the PPTX

Read the pptx skill at /home/node/.claude/skills/pptx/SKILL.md and its pptxgenjs.md reference for the full API and gotchas. Build with PptxGenJS as editable native elements. Output must be an editable widescreen 16:9 .pptx. Do not create HTML, web presentations, screenshots of slides, or rasterized full-slide composites. Generated images may be inserted only as supporting photographic/visual assets.

Core style constants — do not deviate

const CYAN   = "08BFEA"; // electric cyan — primary motion light
const ORANGE = "FF7317"; // saturated orange — counter motion light
const PETROL = "07353C"; // deep petrol — depth panels / plates
const INK    = "111111"; // near-black — type on light canvases
const WHITE  = "FFFFFF"; // paper canvas + thin type on images
const PAPER  = "F5F5F3"; // cool paper gray — alternate canvas
const DISPLAY = "Arial";  // geometric sans display titles — LIGHT / regular weight only
const SANS    = "Arial";  // labels, body, metadata, chart text

Fonts are chosen from the pptx skill's QA-safe list so overflow checks stay trustworthy. Evoke the clean geometric-sans feel through light/regular weight and letter-spacing (charSpacing), never heavy bold — never place a heavy bold title on an image slide. PptxGenJS has no true gradient fill; build any cyan→orange color-field plate from a generated/uploaded image or by layering 2–3 translucent shapes (transparency:), never a baked-alpha hex.

Typography sizing (strong hierarchy, generous space)

Layout — LAYOUT_WIDE (13.3" × 7.5") for a true 16:9 canvas

Set pres.layout = "LAYOUT_WIDE" before adding slides. All coordinates below assume a 13.3 × 7.5 canvas.

The Cover — Follow This Composition Strictly

A premium motion-poster cover. Build every text/rule/thumbnail element as editable native shapes/text over the background image — never an image of a finished cover.

Background:  ONE full-bleed 16:9 cinematic editorial image, cropped to all four edges.
             Reads immediately as electric CYAN vs warm ORANGE motion (anonymous human/topic motion,
             lateral movement, blurred silhouettes). Dark wash no stronger than 12% — keep cyan/orange bright.
             NOT an empty hallway, architecture-only, cosmic, data tunnel, gradient, single portrait,
             dashboard, product UI, sci-fi laser, neon grid, glass shards, or readable faces/text/logos.
Title:       ONE large thin WHITE sans title on ONE line, ~64–76 pt, centered near 42% slide width / 36% height.
             Do NOT merge the product/company name into it (unless the user asks for that exact title).
Wordmark:    product/company name (if any) as a TINY wordmark, top-right or another subtle metadata spot.
Vertical meta: small vertical year/index/metadata near the FAR-LEFT edge, ~12% of slide height.
Thumbnail:   exactly ONE small floating 4:3 thumbnail below & slightly left of the title, ~12–14% slide width,
             a tight cyan-orange crop; subtle translucent backing, no thick frame.
Glyph row:   one tiny row of media-control-style glyphs beside the thumbnail.
Descriptor:  one short two-line descriptor below the thumbnail.
Context:     tiny contact/context details at bottom-left when appropriate.
Hairline:    at most ONE short hairline — never a long rule across the whole slide.

Keep all cover copy under 18 visible words (excluding contact/context details). Do not put a subtitle, tagline, email-as-headline, agenda, metrics, product screenshot, portrait card, KPI, chart, panel, border, or explanatory paragraph on the cover. If elements start to crowd, remove the descriptor or glyph row first.

Content Slides — From Slide 2 Onward

Use a restrained editorial rhythm: alternate white / cool-paper-gray canvas slides with full-bleed motion-photo slides. Build from these modules (all editable native elements except the photos themselves), varying them across the deck: