qwen-delegate

Delegate a task directly to the Qwen CLI (cloud-backed, OAuth-authenticated). Executes the task via the qwen CLI, validates the result, and returns it to the user.

Use this agent when the user asks "what could I delegate to Qwen?", "which tasks can Qwen handle?", "help me save tokens", "analyse this conversation for delegation opportunities", or when the user explicitly asks to identify tasks suitable for the Qwen CLI. Also use proactively after a long task list is presented and token economy is a concern. <example> Context: User has just listed several tasks to complete and is concerned about token usage. user: "I have these 8 tasks to complete today — can any of them go to Qwen?" assistant: "I'll use the qwen-advisor agent to analyse which tasks are good delegation candidates." <commentary> User explicitly wants to identify delegation opportunities across a task list — qwen-advisor is the right tool. </commentary> </example> <example> Context: User wants to understand the cost/benefit of using the Qwen CLI. user: "What kind of tasks should I delegate to Qwen to save tokens?" assistant: "I'll use the qwen-advisor agent to explain delegation strategy and give concrete examples." <commentary> User is asking for guidance on delegation strategy — qwen-advisor provides that analysis. </commentary> </example> <example> Context: User has described a complex workflow with many steps. user: "Review this workflow and tell me which steps Qwen could handle vs which need Claude." assistant: "I'll use the qwen-advisor agent to triage the workflow by delegation suitability." <commentary> Multi-step workflow triage is exactly the qwen-advisor's purpose. </commentary> </example>

This skill should be used when Claude is about to perform tasks that can be delegated to the Qwen CLI to save Anthropic PRO tokens. Delegation candidates include "summarise this", "translate this text", "format this content", "generate boilerplate", "create a template", "write this function", "generate this code", "refactor this snippet", "convert this format", "explain this concept", or any text-processing or code-generation task that does not require Claude's reasoning depth. Always apply a validation gate — Claude reviews Qwen output before presenting to the user.