boyzcl

当用户要让任意具备图像生成能力的 Agent 处理“设计可用性”要求较高的图像任务时使用。这个 skill 会先对齐交付物合同，再通过 information reliability gate、representation strategy、delivery viability gate 与 outcome accountability 把结果推向真正可用的视觉交付。

Use this skill for complex, open-ended, multi-step, or underspecified tasks that should not be executed immediately. Trigger when the user asks the agent to clarify requirements first, asks for one-question-at-a-time alignment, needs explicit requirement consensus before execution, wants a task-specific Markdown execution prompt or document, or wants a full clarification, execution, and self-review workflow. 默认用于复杂、开放式、需求不完整、多阶段、误解成本高的任务：先逐轮单问澄清需求，达到高置信理解后复述并确认，再为当前任务生成一次性 Markdown 执行指令文档，随后执行、自检和总结。 Do not use for simple, well-scoped requests that can be executed directly.

Use this skill when an agent needs to establish, audit, repair, or govern a project's documentation system for human and AI collaboration: document entrypoints, authority or source-of-truth boundaries, current rules vs history vs evidence separation, bounded pilot adoption, retrofit paths for existing projects, document-type templates, and lightweight docs checks. This is a documentation governance skill, not a general writing helper. Use it when docs are messy, duplicated, hard to route, or when agents keep reading the wrong files. Do not use it for ordinary one-off writing or polishing a single document.