run-gemini
向助手发出协助请求，适用于长文阅读理解、网络搜索、网页读取、文章翻译等任务场景。
Security Assessment
About run-gemini
The 'run-gemini' skill allows users to request assistance from an external assistant for tasks like long document reading comprehension, web searches, web page reading, article translation, and even front-end interface development. It leverages the MCP tool to execute tasks within an isolated runtime environment, ensuring that the assistant's actions are free from previous context, providing a clean and focused environment for task execution. This skill is particularly beneficial for complex tasks that require external expertise or for situations where context isolation is crucial for accurate task execution.
FAQ
How do I use the 'run-gemini' skill?
You can use the 'run-gemini' skill by calling the 'mcp__plugin_headless-knight_runCLI__gemini' tool with a set of parameters, including 'prompt', 'systemPrompt', 'workDir', 'model', and 'env'. The 'prompt' is a mandatory parameter that contains the task description, and the others are optional depending on your needs.
Which models are available for use with 'run-gemini'?
There are three models available: 'gemini-3-pro-preview' for complex tasks like long document comprehension and web design, 'gemini-2.5-flash' for network searches and creative writing, and 'gemini-2.5-flash-lite' for rapid responses and brainstorming.
What environment variables are used in the 'run-gemini' skill?
The key environment variables used are 'GEMINI_API_KEY', 'HTTP_PROXY', 'HTTPS_PROXY', 'ALL_PROXY', 'GEMINI_CLI_COMMAND', and 'NODE_ENV'. These are automatically configured if not otherwise specified.
Can I customize the assistant's behavior with a system prompt?
Yes, you can pass a 'systemPrompt' to constrain the assistant's behavior according to specific guidelines or requirements, making it more suitable for the task at hand.
Install run-gemini
Quick Setup:
- Copy the skill folder to
.claude/skills/ - Claude will automatically detect and use the skill
Repository
LostAbaddon/HeadlessKnight