MCP App StoreMCP App Store
from bananahub-skill117
Multi-provider image workflow that normalizes and optimizes non-English prompts, generates and edits images across Gemini/OpenAI providers, and manages reusable
BananaHub adds a consolidated image-generation and editing workflow to an agent. It normalizes mixed-language or non-English prompts into English, applies conservative prompt optimization, routes requests to provider adapters (Gemini/Nano Banana, OpenAI, or compatible endpoints), and supports template-based generation and discovery. The skill handles generation, mask edits, multi-reference edits, and template-driven workflows while offering model recommendations and safe fallbacks.
Use BananaHub when a user explicitly requests /bananahub or mentions BananaHub and needs image generation or editing from multiple provider backends — especially for non-English prompts or when template-driven, repeatable workflows are desired. Do NOT trigger on generic image-generation messages like "生成图片". Typical triggers: /bananahub, bananahub discover, bananahub edit, and localized phrases calling out BananaHub.
Best suited for Claude Code and other agent runtimes that can call Python scripts and supply provider API keys; can also be adapted for systems using Open Agent Skills that run external skill scripts.
BananaHub is a multi-provider image generation workflow skill with prompt optimization, template discovery, and model fallback. 2 of 4 scripts failed on import due to intra-package module resolution (providers/ subpackage not on PYTHONPATH in isolated test); the other 2 ran clean. The SKILL.md is comprehensive but very long. Telemetry phones home to worker.bananahub.ai (opt-out via env var). Auto pip-install of missing deps guarded by interactive prompt.
providers (module - part of skill's own package, not standalone)Telemetry to external host (worker.bananahub.ai) with anonymous ID, but respects BANANAHUB_DISABLE_TELEMETRY and only sends limited event types (selected/generate_success/edit_success). No hardcoded secrets. subprocess.run only used for pip install with user confirmation in interactive mode. No shell injection patterns found. No os.system/eval/exec calls. Architecture is well-structured with clear separation between runtime config, config store, providers, and main CLI.
Bounty Hunter — Atlas
Persona skill: 'Atlas' — a profit-focused developer persona for discovering, evaluating and executing paid bounties or freelance tasks with ROI-aware workflows.
FastAPI Project Templates
Creates production-ready FastAPI project scaffolds with async patterns, DI, middleware, and testing best practices for high-performance APIs.
Generate Multilingual Sentences with Contextual Definitions
Produce multiple example sentences using a target word in varied meanings, translate them into a target language, and show the contextual meaning of the word fo
AI News Daily
Generate a structured, bilingual daily briefing of the latest AI industry news, model updates, papers, and open-source tools with in-depth technical analysis an
Prompt Engineering Patterns
A practical collection of prompt engineering techniques and patterns for building reliable, controllable LLM prompts: CoT, few-shot, structured outputs, templat
Confluence Expert
Manage Confluence spaces, templates, permissions and documentation workflows; create page hierarchies, macros and governance for team knowledge bases.
Create & Configure Agents
Helpers and templates to scaffold agent directories, generate agent.yaml, and author SOUL.md and RULES files for new agents.
DSPy — Declarative LM Programming
Use DSPy to build declarative, modular LM pipelines, optimize prompts automatically, and assemble reliable RAG/agent systems with structured signatures and opti