Claude Cowork文件夹系统提示方案

1. Folder: A folder on your computer, "Claude Cowork". Every file inside replaces a prompt you'd type. No more "here's the context again" in every chat. 2. ABOUT ME/: about-me.md & anti-ai-writing-style.md file. Who you are. How you write. What you'd never say. 80% of these files should be what you REJECT. Claude reads them before every task. 3. PROJECTS/: One subfolder per project. Inside: brief.md, references/, drafts/. Prompt: "Read everything in PROJECTS/[name] before starting." Live work lives here. Stop pasting 15 context messages. Put them in files. 4. TEMPLATES/: Save your best work as reusable patterns. Prompt: "Use the template in TEMPLATES/[name]." Claude studies the structure before creating new. Add: "Match the length, sections, and tone. Only replace the content." 5. CLAUDE OUTPUTS/: The ONLY folder Claude writes to. One subfolder per project, mirroring PROJECTS/. You always know where deliverables live. No more "Untitled 47.docx" on your screen. 6. Global Instructions: Set it once. It runs every single time: Settings -> Cowork -> Edit Global Instructions. Folder protocol, naming conventions, operating rules. You tell Claude how to behave ONCE. 7. The One Prompt: Your new prompt is one sentence: "I want to [TASK] so that [SUCCESS CRITERIA]. Start by using AskUserQuestion." 80% of my chats start with exactly this. Claude prompts YOU. Not the other way around. 8. Obsidian (free): Open your Cowork folder as a knowledge base. Your .md files start looking like Google Docs. You can edit, search, and link them in seconds. Your brain + Claude's memory, in one clean place. 9. Alignment: The folder does the talking. The prompt is a line. Claude asks YOU questions before doing anything. This replaces the old prompting era.