v0.0.3

What's Changed

• Release 0.0.2 — Landing site, design system, and AI discovery by @bytonylee in #1
• Rename GPT image slide skills by @bytonylee in #3
• Refactor site to static deployment by @bytonylee in #4
• Add tightened-slide skill workflow by @bytonylee in #2

New Contributors

• @bytonylee made their first contribution in #1

Full Changelog: 0.0.2...v0.0.3

0.0.2 — Landing site, design system, and AI discovery

0.0.2 — Landing site, design system, and AI discovery

This release ships https://future-slide.tonylee.im — a complete
landing page for the future-slide skill bundle, plus the design
documentation and AI/SEO scaffolding that supports it.

The four skill bundles (gpt-slide-design, gpt-slide-plan,
gpt-slide-prompt, gpt-slide-generate) are unchanged in behavior.
This release is additive — install the same way as 0.0.1.

Highlights

• 🌐 New landing page at https://future-slide.tonylee.im
• 🇰🇷 i18n with English (default) and Korean, auto-detected from the
  browser and persisted in localStorage
• 🌓 Light + dark themes (light by default), with a pre-paint script
  that prevents flash on first load
• 🎨 Editorial layout aligned with tonylee.im: narrow max-w-2xl
  column, Anton uppercase wordmark, Source Serif 4 section labels,
  Pretendard for Korean, Phosphor icons throughout
• 🤖 Full SEO + AEO surface: canonical, hreflang, JSON-LD @graph,
  /llms.txt, /robots.txt (welcomes GPTBot / ClaudeBot /
  PerplexityBot / Google-Extended / Applebot-Extended / CCBot),
  /sitemap.xml
• 📐 New site/DESIGN.md documenting the layout, typography, color,
  iconography, and component recipes

Site stack

• Vite 6 + React 19 + strict TypeScript project references
• Tailwind CSS v4 via @tailwindcss/vite, with the Urban Loft
  tokens bridged to Tailwind via @theme inline
• Bun as package manager, runtime, and lockfile
• @phosphor-icons/react for iconography (regular weight, matching
  tonylee.im)

Design system

• Wordmark: Anton, uppercase, clamp(2.25rem, 6vw, 3.75rem)
• Section label (h2): Source Serif 4, clamp(1.2rem, 2.2vw, 1.55rem),
  letter-spacing: -0.04em
• Body: Barlow with letter-spacing: -0.02em
• Korean: Pretendard Variable, word-break: keep-all,
  letter-spacing: 0
• Primary brand: terracotta #a35e47 (light) / #c97d63 (dark) —
  used only for selection, focus rings, and the hero forward link
• Vignette body: radial gradient fading to neutral-100 (light) /
  black (dark), background-attachment: fixed desktop and scroll
  on coarse pointers

SEO + AEO

• Title format: future-slide | Disciplined slide generation for Codex
• Meta description and Open Graph mirror the title format
• Hreflang for en, ko, x-default all pointing at the canonical
  URL (single-page app handles locale via JS)
• Twitter / X card with @jeongmin1604
• JSON-LD graph: WebSite → SoftwareApplication (Apache-2.0,
  free Offer, code repository) → Person (Tony Lee) → FAQPage
  with four answers (what is future-slide, why four skills, how to
  install, what artifacts the pipeline produces)
• /llms.txt mirrors the structure used on tonylee.im: project
  summary, four-skill workflow, install instructions, artifact
  chain, citation format, AI-crawler welcome
• /robots.txt allowlists every major AI crawler explicitly

Page structure

[ Floating header ] tony · GitHub · 🌐 · 🌓
[ Hero ]
   future-slide                                           ← Anton h1
   ⏱ v0.0.2  |  ⚖ Apache-2.0
   Disciplined slide generation, one stage at a time.
   ...lede...
   [GitHub] View skill / 스킬 보러가기 →                    ← combined CTA
[ Flow ] locale-aware four-stage diagram (PNG)
[ Workflow ] 01 - 04 editorial list
[ Why four skills ] quiet bullets
[ Examples ] terminal codeblock + artifact chain
[ Install ] CLI + manual codeblocks
[ What it optimizes for ] quiet bullets
[ Footer ] Apache-2.0 · GitHub · © 2026 Tony

Quality

• Em dashes (—) removed from every visible UI string and
  /llms.txt so Korean line-breaking stays natural
• Pre-push security audit: no secrets in source, lockfile, or
  build outputs; .env.example is the only env file committed
• .gitignore covers Vite outputs, env files, oh-my-claudecode
  runtime state (.omc/), and IDE/OS metadata
• Ships with a clean bun run build (244 KB JS / 75 KB gzip,
  24 KB CSS / 6 KB gzip, 8 KB HTML / 2.5 KB gzip)

Documentation

• site/DESIGN.md (~310 lines): layout, responsive rules,
  typography roles, color tokens, Phosphor icon catalog,
  component recipes (tag pill, list item, codeblock, forward
  link, primary CTA), hover-dim behavior, accessibility, and
  the page outline
• README badges trimmed to Version + License (the Codex Skills
  shield is removed in both English and Korean)

What's next

• Rolling the per-skill SKILL.md files into the landing page
  workflow section so install + reference stay in one place
• Adding programmatic SEO pages for popular deck types
  (research summary, partnership memo, equity report) once the
  workflow stabilizes
• Connecting the production deploy on Vercel and pointing
  future-slide.tonylee.im at it

0.0.1

Future Slide Skill 0.0.1

Initial public release of the Future Slide skill bundle for Codex.

This release packages a disciplined slide-generation workflow that separates design extraction, deck planning, page-level prompting, and sequential image generation. The goal is to keep generated decks consistent across pages instead of letting each slide drift into a different generic layout.

Highlights

• Adds a staged workflow for creating slide decks from reference visuals, user files, and a deck request.
• Extracts reusable slide design rules into DESIGN.md before planning or generation begins.
• Plans deck structure in JSON before writing page-level prompts.
• Writes strict per-slide prompt JSON with layout, hierarchy, table, chart, icon, infographic, and footer constraints.
• Generates final slide images one page at a time with stable page-numbered filenames.
• Includes a design-system template and a visual workflow diagram.

Included skills

future-slide

Provides the orchestration contract for the workflow. It enforces the order of design extraction, deck planning, and page-prompt generation so the model does not skip directly into slide writing before the visual system is defined.

gpt-slide-design

Converts reference slide images into a reusable DESIGN.md.

The skill focuses on visual language rather than slide content: palette, typography, spacing, headers, body zones, footers, chart/table treatment, icon behavior, infographic cards, diagram flow, and repeatable body-slide rules.

gpt-slide-plan

Builds the deck logic before page prompts are written.

It uses DESIGN.md, user files, and the user's goal to choose page order, narrative flow, evidence placement, page families, and layout families. It is designed to avoid simple file-order summaries and instead produce a persuasive deck structure.

gpt-slide-prompt

Turns DESIGN.md and slide_plan.json into detailed page-by-page prompt JSON.

Each slide prompt describes the objective, narrative function, content blocks, layout zones, design constraints, content constraints, and generation notes so downstream image generation has less room to improvise generic slide layouts.

gpt-slide-generate

Renders slide images sequentially from DESIGN.md and slide_prompts.json.

It requires page-by-page generation, visual inspection, and saving outputs into the project with stable names such as page_1.png, page_2.png, and page_N.png.

Documentation and assets

• Adds the main README with workflow explanation, examples, expected artifacts, and installation pattern.
• Adds templates/DESIGN_TEMPLATE.md as a scaffold for consistent design extraction output.
• Adds public/four-skill-flow.png as the workflow diagram used by the README.
• Adds repository ignore rules for local runtime state, caches, and OS metadata.

Typical workflow

$gpt-slide-design [reference slide image]
$gpt-slide-plan /path/to/report.pdf Create a research-summary deck.
$gpt-slide-prompt
$gpt-slide-generate

Expected artifact chain:

1. DESIGN.md
2. slide_plan.json
3. slide_prompts.json
4. page_1.png ... page_N.png

Release contents

The 0.0.1 tag points to the initial packaged skill bundle commit. Later documentation polish, including Korean README, Apache-2.0 license file, navigation links, and expanded installation guidance, is available on the main and dev branches after this tag.

Notes

This is the first release, so the focus is on packaging the skill workflow, documentation, and reusable prompts. There is no application runtime or automated test harness in this repository yet.