Add MythosForge image-generation skill (Replicate: flux-schnell / nano-banana-2 / retro-diffusion)

[ryjin111]

MythosForge image generation

Adds mythosforge/ — a skill that generates an image from a text prompt via Replicate. It's the image-generation engine MythosForge uses in production, packaged for Bankr agents.

Pick the engine with --model:

--model	Replicate model	Best for	Formats
flux-schnell ⭐	black-forest-labs/flux-schnell	fast, general (default)	webp/png/jpg
nano-banana-2	google/nano-banana-2	high-quality, general (2K)	png/jpg
retro-diffusion	retro-diffusion/rd-plus	pixel-art / game assets	png

Usage

export REPLICATE_API_TOKEN=r8_...
./scripts/generate.sh "a heroic fox knight in golden armor" -a landscape -o hero.png
./scripts/generate.sh "pixel-art knight with a sword" -m retro-diffusion -o knight.png

Files

• SKILL.md — skill definition (clawdbot metadata, usage, model table)
• references/image-gen-api.md — per-model Replicate input schemas + prediction lifecycle
• scripts/generate.sh — bash/curl/jq; fail-closed (missing token, non-2xx, failed/canceled prediction, timeout, or empty download all exit non-zero and write no partial file)

Tested

Live-tested all three models end-to-end against Replicate — each returned a real, correctly-sized image:

• flux-schnell → 1344×768 PNG
• nano-banana-2 → 2752×1536 PNG (2K)
• retro-diffusion → 256×256 PNG (pixel art)

Requires a REPLICATE_API_TOKEN.