向助手发出协助请求,适用于需要在隔离的独立环境中执行复杂编程任务、任务规划、文稿撰写的场景。
The 'run-claude' skill allows users to delegate complex tasks such as programming, task planning, document writing, and even creative works to an external assistant in an isolated and clean runtime environment. This skill is ideal for scenarios where tasks need to be executed in a secure, compartmentalized environment to ensure task isolation from the current context. It leverages the MCP tool to send requests to an external assistant and provides the flexibility to choose from various AI models tailored to specific task requirements.
The key features of 'run-claude' include the ability to pass specific parameters such as task descriptions, system prompts, working directories, model types, and custom environment variables. It supports models like 'sonnet', 'haiku', and 'opus', each designed for different use cases. For example, the 'sonnet' model excels at complex programming tasks and writing various types of content, while 'opus' is suited for deep analysis and professional document creation. The flexibility to specify environment variables like API keys and proxy settings ensures the tool can be adapted to various workflows. This skill is particularly useful for developers, content creators, and anyone needing an AI assistant for detailed, high-precision tasks.
Use cases for this skill span multiple domains including software development, creative writing, marketing copy generation, and even legal document drafting. Its ability to handle complex and high-stakes tasks in a dedicated, isolated environment makes it an excellent choice for professionals who require both precision and security in their workflows.
You can use the 'run-claude' skill by invoking the `mcp__plugin_headless-knight_runCLI__claude` tool, providing parameters like 'prompt' for the task description, 'systemPrompt' for constraints, 'workDir' for the working directory, 'model' for the AI model selection, and 'env' for any custom environment variables.
There are three models available: 'sonnet' for complex programming tasks and content creation, 'opus' for deep analysis, legal documents, or high-precision writing, and 'haiku' for creative arts and brainstorming. Choose the model based on your task's complexity and type.
By default, the skill will use the environment settings from the current context. However, you can set custom environment variables such as 'ANTHROPIC_API_KEY', 'HTTP_PROXY', 'HTTPS_PROXY', and 'NODE_ENV' to tailor the skill to your needs.
The main limitation is that the skill relies on external infrastructure for task execution, which means it may require access to specific resources like API keys or proxies. Also, tasks are isolated from the current context, which might limit interactions with other tools or ongoing processes.
Yes, if no model is specified, the default model 'sonnet' will be used. However, it is recommended to choose the model that best fits your task for optimal performance.
Quick Setup:
.claude/skills/Repository
LostAbaddon/HeadlessKnight