Nano Banana Pro Prompt Recommender Skill
AI agent skill that searches 10,000+ curated Nano Banana Pro (Gemini) image prompts and recommends the best matches in one sentence.
Evidence shows: repository includes automated update workflow using GitHub Actions secrets (CMS_HOST, CMS_API_KEY) with contents: write permission, no excessive permissions found. Install script (setup.js) not seen in static review, but postinstall executes, posing potential risk. Data flow transparency: README states data sourced from CMS, but does not detail how user data is handled. Sensitive data handling: no mention of user data collection or processing. Dependency security: few dependencies (dotenv, qs-esm), but no vulnerability scanning evidence. External effects: automated updates modify repository content, but expected behavior. Rollback: no rollback mechanism provided. Source attribution: README claims data from YouMind community, but no specific source links. Deductions: no user confirmation mechanism, opaque data flow, missing sensitive data handling, missing rollback.
Evidence shows: README and package.json descriptions consistent, SKILL.md not provided but README mentions its existence. Dependency availability: few dependencies, but no lock file evidence (package-lock.json or pnpm-lock.yaml). Failure messages: no documentation of error handling or failure messages. Deductions: dependency availability not fully verified, failure messages missing.
Evidence shows: README clearly identifies target users (AI assistant users) and use cases (search, remix), provides multi-language support. Capability boundaries: README states support for multiple AI assistants, but no explicit limitations. Trigger precision: provides example queries, but no precise trigger words defined. Environment fit: provides multiple installation methods, but no environment requirements (e.g., Node version). Deductions: capability boundaries and trigger precision not explicit enough.
Evidence shows: README well-structured with installation, usage, FAQ, project structure. Installation notes detailed with multiple methods. Naming stability: version number exists, but no changelog. Examples and FAQ: multiple examples and FAQ provided. Known limitations: not explicitly listed. License: MIT license, but LICENSE file content not provided. Versioning: package.json has version, but no changelog. Maintenance responsibility: author is YouMind-OpenLab, but no maintenance commitment. Deductions: missing changelog, known limitations, LICENSE file content.
Evidence shows: Output usability: recommendations include title, description, prompt, sample images. Marginal value: provides 10,000+ prompts and remix mode, unique value. Cost-benefit: free to use, but no performance data. Deductions: cost-benefit not fully demonstrated.
Evidence shows: README claims 10,000+ prompts, but no specific data source or verification method. Cross-source corroboration: no other sources to verify. Fact-inference separation: README distinguishes facts (prompt count) and inferences (quality), but not explicitly labeled. Deductions: lack of verifiable data source and cross-source verification.
- Install script (setup.js) not seen in static review, but postinstall executes; review its content for safety.
- Automated update workflow uses GitHub Actions secrets; ensure secrets are securely stored to avoid leakage.
- Data sourced from CMS, but no specific source links; verify data authenticity and copyright.
What does this agent do, and when should you use it?
This AI agent skill gives assistants like Claude and OpenClaw the ability to intelligently search and recommend from a library of 10,000+ curated Nano Banana Pro (Gemini) prompts, categorized by use case and featuring sample images. It supports semantic search and a content remix mode for personalized prompts. The skill is token-efficient, using grep-style search that never loads full files, and the library updates twice daily. It response in multiple languages but always provides English prompts for generation. Installation is available via ClawHub or npx for OpenClaw and Claude Code.
The skill identifies the prompt category from the user's need (e.g., social media, product marketing, avatar) and searches local JSON files (e.g., social-media-post.json) returning up to 3 matches with translated titles and descriptions, the exact English prompt, sample images, and whether reference images are needed. In remix mode, users paste content and the skill recommends style templates, asks personalizing questions, and generates a custom prompt. The flow: user describes need → skill identifies category → searches files → returns recommendations → optional remix.
- Content creators need eye-catching image prompts for Instagram posts.
- Marketing professionals need product image prompts for ads or campaigns.
- Social media managers are looking for YouTube thumbnail ideas for tech review videos.
- E-commerce sellers require white-background product photos for listings.
- Podcast hosts want cover illustrations based on episode scripts.
What are this agent's strengths and limitations?
- Over 10,000 curated prompts organized by use case with sample images.
- Smart semantic search allows natural language descriptions.
- Content remix mode generates custom prompts based on user content.
- Twice daily updates keep the library current with viral prompts.
- Multi-language responses with English prompts for generation.
- Prompts are optimized for Nano Banana Pro and may need adjustments for other models.
- Data source relies on community shares, subject to platform policies.
- Image generation requires separate Gemini access or YouMind service.
- Requires Node.js 20+ and pnpm, which may be a barrier for non-technical users.
- License is MIT but unclear if the data is fully open-sourced.
How do you install or deploy this agent?
Installation varies by assistant: for OpenClaw, run clawhub install nano-banana-pro-prompts-recommend or search in chat; for Claude Code, run npx skills i YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill; for other assistants like Cursor, Codex, use the universal installer npx skills i YouMind-OpenLab/nano-banana-pro-prompts-recommend-skill; manual install via npx openskills install is also possible.
How do you use this agent?
Mode 1: Direct search — describe what you need, e.g., "Find me a cyberpunk-style avatar prompt", and get up to 3 recommendations with translated titles, English prompts, and sample images. Mode 2: Content remix — paste your article or script, and the skill will recommend style templates, ask personalizing questions, and generate a custom prompt tailored to your content.
How does this agent compare with similar options?
No clear comparisons provided.