Nano banana korean rendering
v1.0.3비라틴 문자(한글, 일본어, 중국어 등)를 AI 이미지에 정확히 렌더링하는 스킬. Canvas 프리렌더링과 Gemini를 활용하여 텍스트 깨짐 없이 이미지를 생성합니다.
⭐ 1· 1.7k·1 current·1 all-time
MIT-0
Download zip
LicenseMIT-0 · Free to use, modify, and redistribute. No attribution required.
Security Scan
OpenClaw
Suspicious
medium confidencePurpose & Capability
The skill's code (render.mjs, setup.mjs) and SKILL.md align with the stated purpose of Canvas pre-rendering + Gemini analysis/generation. However, the registry metadata lists no required environment variables or primary credential while SKILL.md and setup.mjs clearly require GEMINI_API_KEY for full functionality — this mismatch is an incoherence in the declared capabilities/requirements.
Instruction Scope
The SKILL.md instructions are focused on detecting scripts, analyzing text with Gemini, rendering PNGs with Canvas, and sending them to Gemini for final generation. The instructions explicitly call out when network access and the GEMINI_API_KEY are used. They do not instruct reading unrelated system files or exfiltrating arbitrary data, but do send user prompts and rendered images to an external service (Gemini).
Install Mechanism
There is no platform install spec, but setup.mjs runs `npm install` (execSync) in the skill directory and downloads font files from GitHub raw URLs. The download hosts are GitHub (well-known). Installing npm deps and writing font files into the skill's fonts directory is expected for this functionality, but running install scripts and child_process is a higher-risk operation and should be executed in a controlled environment.
Credentials
The registry declares no required env vars, yet SKILL.md marks GEMINI_API_KEY as required and setup.mjs checks for it; the skill will use that key to call Gemini (analysis and image generation). Requesting a single service API key for the service it integrates with is proportionate, but the omission from declared metadata is an inconsistency that reduces transparency and should be corrected before installation.
Persistence & Privilege
The skill does not request always:true and does not modify other skills. Its setup writes node_modules and a fonts directory under its own folder. It declares trigger keywords that allow the agent to auto-activate when prompts contain non-Latin text — autonomous invocation is the platform default, so this is expected but worth noting because it causes the skill to run when such prompts are seen.
What to consider before installing
What to consider before installing:
- Credential mismatch: The registry metadata does not list any required env vars, but SKILL.md and setup.mjs require GEMINI_API_KEY. Confirm this omission with the publisher and only provide a key you trust and can rotate. Prefer a scoped/tokenized key with minimal privileges.
- Network & data flow: The skill will send prompts and pre-rendered PNGs to Google Gemini (via @google/generative-ai). Do not use it with sensitive or private text/images unless you accept that data will be transmitted to that external service.
- Installation side effects: setup.mjs runs npm install (execSync) and downloads font files from GitHub raw URLs, writing node_modules and fonts into the skill's directory. Run setup in an isolated/sandbox environment or inspect the exact npm install output before trusting it.
- Code review: Although the visible code appears to implement the stated pipeline, review render.mjs fully (the provided content was truncated) to ensure there are no unexpected network endpoints, telemetry, or file accesses beyond fonts/node_modules and Gemini SDK calls.
- Safety steps: run the setup and skill in a sandbox/container, verify the fonts come from trusted sources, set GEMINI_API_KEY in a secure manner (not embedded in files), and monitor outgoing network connections during initial use.
If you require help with specific checks (full code audit, network monitoring steps, or how to run in a sandbox), I can guide you through them.Like a lobster shell, security has layers — review code before you run it.
latestvk97bf4ajzvh3g36ne5f6rzdfp580qvpn
License
MIT-0
Free to use, modify, and redistribute. No attribution required.
