mofa-skills

AI-powered content generation platform that turns text into visual media — presentations, greeting cards, comics, infographics, and animated videos — using Google Gemini and Alibaba Dashscope APIs.

Skills

Skill	Output	Styles	Description
mofa-slides	.pptx	14	Presentation decks with full-bleed AI images and editable text overlays
mofa-cards	.png	7	Greeting cards in Chinese art styles (ink-wash, guochao, etc.)
mofa-comic	.png	5	Multi-panel comic strips (xkcd, manga, ligne-claire, etc.)
mofa-infographic	.png	4	Multi-section infographics stitched vertically
mofa-video	.mp4	4	Animated video cards with background music via Gemini Veo
mofa-research	text	—	3-agent deep research pipeline (search → analyze → synthesize)
mofa-videolizer	.srt	—	Subtitle generation from text + audio (Whisper / basic fallback)
mofa-workflow	artifacts	—	Multi-agent team pipeline (architect → developer → reviewer → tester)
mofa-news	.md	8 categories	News digest from Google News, Hacker News, Yahoo, Substack, Medium
mofa-github	text	—	GitHub integration via gh CLI (issues, PRs, CI, releases, API)
mofa-public-apis	text	40+ categories	Search free public APIs — browse by category, auth, HTTPS, CORS
mofa-mcdonalds	text	—	McDonald's China ordering via MCP — 点餐、领券、积分兑换
mofa-xhs	text	—	Xiaohongshu (小红书) integration — search, read, like, comment, publish
mofa-crawler	text	—	Web crawling via Cloudflare Browser Rendering API — full-site extraction with JS rendering
mofa-logo	SVG	8 styles	AI logo generation with Claude Opus 4.6 — minimalist, mascot, emblem, wordmark
mofa-fm	.wav	—	Voice TTS with custom voice cloning via Qwen3-TTS on Apple Silicon
mofa-site	static HTML / starter source	Quarto, Astro, Next.js, React	Build lesson sites, scaffold extracted templates, and run a server-backed site studio
mofa-publish	live URL	GitHub Pages, Mac Mini	Deploy built static sites and verify the live URL

DingTalk Wukong Skills (from stvlynn/dingtalk-wukong-skills)

Skill	Description
docx	Advanced Word document creation/editing with tracked changes, comments, formatting preservation
xlsx	Excel spreadsheet processing with formulas, data analysis, and visualization
pptx	PowerPoint presentation creation/editing, including HTML-to-PPT conversion
pdf	PDF processing — text/table extraction, creation, merging/splitting, form filling
pdf-convert-to-word	Professional PDF to Word/Markdown conversion
dws	DingTalk workspace operations (OA, calendar, docs, todo, chat, approval, attendance, etc.)
12306-train-query	Real-time 12306 train ticket and schedule queries
ctrip-flight-search	Global flight search and price comparison via Ctrip
dianping-info-query	Local merchant info and reviews from Dianping (大众点评)
skill-creator	Meta-tool for designing, validating, and packaging new agent skills

Architecture

mofa-skills/
├── mofa/                 # Shared engine (Node.js) + config
│   ├── lib/
│   │   ├── engine.js           # Image generation orchestration
│   │   ├── toml-style.js       # Style file loader
│   │   └── image-providers.js  # Gemini & Dashscope API clients
│   └── config.json             # API keys + defaults
│
├── mofa-slides/          # 14 presentation styles
├── mofa-cards/           # 7 greeting card styles
├── mofa-comic/           # 5 comic strip styles
├── mofa-infographic/     # 4 infographic styles
├── mofa-video/           # Video animation styles
├── mofa-research/        # DOT-based research pipeline
├── mofa-research-2.0/    # DeerFlow + mofa-research hybrid
├── mofa-videolizer/      # Subtitle generation (Whisper / basic)
├── mofa-workflow/        # DOT-based multi-agent team pipeline
├── mofa-news/            # News digest aggregator
├── mofa-github/          # GitHub integration via gh CLI
├── mofa-public-apis/     # Public API discovery (local cache)
├── mofa-mcdonalds/       # McDonald's China ordering via MCP Server
├── mofa-xhs/             # Xiaohongshu (小红书) integration via xhs-cli
├── mofa-crawler/         # Web crawling via Cloudflare Browser Rendering API
├── mofa-logo/            # AI logo generation with Claude Opus 4.6
├── mofa-fm/              # Voice TTS + cloning (Pure Rust, via ominix-api)
├── mofa-site/            # Multi-template website builder and studio
├── mofa-publish/         # Static site deployment to GitHub Pages / Mini hosting
│
└── mofa-cli/             # Pure Rust CLI (single binary, no Node.js)
    └── src/
        ├── main.rs             # CLI entry (slides|cards|comic|infographic|video)
        ├── gemini.rs           # Gemini API client
        ├── dashscope.rs        # Qwen-Edit client
        ├── layout.rs           # VQA text extraction + font calibration
        ├── pptx.rs             # PPTX builder (DrawingML XML)
        ├── image_util.rs       # Image stitching
        ├── veo.rs              # Veo video generation
        └── pipeline/           # Per-skill generation pipelines

Two implementation stacks share the same config format, style system, and TOML templates:

• JavaScript engine (mofa/) — Node.js, requires @google/genai, pptxgenjs, sharp
• Rust CLI (mofa-cli/) — single binary, zero Node.js dependency

Setup

Prerequisites

• GEMINI_API_KEY — required for all skills
• DASHSCOPE_API_KEY — optional, for Qwen-Edit image refinement
• Node.js — for the JS engine
• ffmpeg — for video compositing (mofa-video)
• ImageMagick (magick) — for comic/infographic stitching (JS engine only)
• Quarto — for mofa-site Quarto lesson builds
• GitHub CLI / SSH — for mofa-publish deployments

Configuration

Copy the example config and set your API keys:

cp mofa/config.example.json mofa/config.json

Edit mofa/config.json — API keys use env:VAR_NAME to read from environment variables:

{
  "api_keys": {
    "gemini": "env:GEMINI_API_KEY",
    "dashscope": "env:DASHSCOPE_API_KEY"
  }
}

Or export them directly:

export GEMINI_API_KEY="your-key-here"
export DASHSCOPE_API_KEY="your-key-here"

JavaScript engine

cd mofa && npm install

Rust CLI

cd mofa-cli && cargo build --release
# Binary at mofa-cli/target/release/mofa-cli

Usage (Rust CLI)

# Presentation deck
mofa slides --style nb-pro --out deck.pptx --slide-dir /tmp/slides --input slides.json

# Greeting cards
mofa cards --style cny-guochao --card-dir /tmp/cards --input cards.json

# Comic strip
mofa comic --style xkcd --out comic.png --input panels.json --layout horizontal

# Infographic
mofa infographic --style cyberpunk-neon --out poster.png --input sections.json

# Animated video card
mofa video --style video-card --anim-style shuimo --card-dir /tmp/video --input cards.json

See each skill's SKILL.md and mofa-cli/README.md for full API documentation.

Style System

Styles are TOML files with prompt variants. Each skill directory has a styles/ folder:

[meta]
name = "nb-pro"
display_name = "NB Pro"

[variants]
default = "normal"

[variants.normal]
prompt = """
Create a presentation slide image...
"""

[variants.cover]
prompt = """
Dark background, centered title...
"""

Adding a new style is as simple as dropping a .toml file into the appropriate styles/ directory.

Models

Role	Default	Used By
Image generation	gemini-3.1-flash-image-preview	All visual skills
Vision QA	gemini-2.5-flash	autoLayout text extraction
Image editing	qwen-image-edit-max-2026-01-16	Text removal refinement

All model names are configurable in mofa/config.json.

Skill Registry

The octos-hub is automatically synced on every push to main after CI passes.

How it works

1. scripts/gen-registry.py scans all mofa-*/ directories for manifest.json or SKILL.md
2. Collects skill names, binary requirements (requires.bins from manifests)
3. Merges with curated metadata from registry-meta.json (tags, description, excludes)
4. The sync-registry CI job pushes the generated registry.json to octos-org/octos-hub via GitHub API

Adding/removing skills

• Add a skill: Create mofa-<name>/ with a manifest.json or SKILL.md. It will appear in the registry on next CI pass.
• Remove a skill: Delete the directory, or add it to exclude_skills in registry-meta.json.
• Update tags/description: Edit registry-meta.json.
• Add binary requirements: Set requires.bins in the skill's manifest.json (e.g., "requires": {"bins": ["ominix-api"]}).

mofa-site Design Principle

mofa-site is static-first.

• The studio can be stateful, but the generated site should build to static assets and deploy cleanly with GitHub Pages or mofa-publish.
• Dynamic site features should default to a shared multitenant backend API, not a dedicated Node/Rust server per sub account.
• A per-site runtime is an exception for advanced cases, not the default website generation model.

Setup (repo admin)

Add a OCTOS_HUB_TOKEN secret to this repo — a GitHub PAT with repo scope for octos-org/octos-hub.

Mini profile deploy

Deploy skills to a specific Mini profile or sub-account:

./scripts/deploy-mini.sh mini1 dspfac
./scripts/deploy-mini.sh mini1 dspfac--newsbot

License

Apache License 2.0