mofa-org/mofa-skills/mofa-cards

mofa-cards

Generate AI-powered greeting / holiday cards as PNG images (full-bleed Gemini artwork in 8 Chinese art styles plus inline custom prompts).

• CLI: mofa cards
• Styles: styles/*.toml (8 built-in + 7 t-shirt variants)
• Config: mofa/config.json

The mofa_cards tool is spawn_only — generation runs in the background and images are delivered when ready. Do NOT wait inside the agent loop.

Load-bearing rules (do NOT skip)

1. Output paths

Use RELATIVE paths. Never prefix skill-output/ yourself — the Octos host rebinds plugin output paths to <workspace>/skill-output/ automatically; a manual prefix double-prefixes and breaks delivery. Never use absolute paths like /tmp/cards/ or /Users/.../Desktop/.

mofa-cards-<YYYYMMDD-HHMMSS>/

2. API key

Verify GEMINI_API_KEY is set before calling the tool. If unset, ask the user to export it — the tool will fail otherwise.

3. Style selection is interactive

Before generating, gather: occasion, style, card count (typically 1-3), aspect ratio (default 9:16). On Telegram, use an inline keyboard for style picks. Recommend a style based on occasion — see docs/styles.md.

4. Inline custom styles are supported

You are NOT limited to built-in styles. Write a full style description directly in the card's prompt field; pick any built-in --style as a layout base. See docs/styles.md for trigger-phrase → snippet recipes.

5. Timeouts preserve cache

Each card is ~15-30s; tool timeout is 600s. If generation times out, cached cards are preserved — rerun and only missing cards regenerate. See docs/examples.md for the timing table.

Where to look next

• Built-in style catalog + occasion mapping + inline custom-style recipes →

docs/styles.md

• Quick-start CLI snippet + input JSON shape + Telegram inline-keyboard example
• timing table → docs/examples.md
• Full CLI flag reference + mofa/config.json keys → docs/cli-flags.md
• Enumerate available styles directly → list styles/*.toml