01
Agent-native first
The skill owns the conversation, constraint extraction, clarification, and iteration. BananaHub should plug into that workflow, not replace it with a separate prompt-builder maze.
About BananaHub
The starter pack and template network for BananaHub Skill.
BananaHub Skill is the agent-native runtime for guided Gemini image generation in chat. BananaHub bundles a small starter pack and adds a searchable, installable, machine-readable layer for everything else.
What BananaHub Is
Not a prompt dump. A distribution layer for reusable templates.
BananaHub exists so BananaHub Skill can ship a tight starter set while reusable prompt and workflow structures remain searchable, installable, and portable across repositories.
02
Progressive disclosure over overload
Users should see only enough structure to make the next good decision: gallery first, template details on demand, full source only when it is needed.
03
Searchable, installable, machine-readable
Templates should be discoverable by humans, installable by CLI, and readable by agents from stable files without scraping the website UI.
Machine Files
Use these files directly for automation.
They mirror the public catalog and are the preferred interface for agents, scripts, and workflow tooling.
Command patterns
/bananahub use cute-sticker
bananahub add user/repo/template
Contribution Loop
A decentralized template network, not a monolithic prompt library.
For template authors
Publish self-describing template folders in GitHub, follow the BananaHub template format, and get indexed through the catalog source. Your modules stay portable in your repo; BananaHub stays lightweight.
Read the template specFor advanced users and agents
Use the catalog to detect a likely match, inspect the source template when needed, and install only the exact module that improves the current task.
Open the bananahub CLI