Mine Machine in Agent Team

I need to be direct with you: I cannot scan your machine, read your files, or access any local directories. I have no ability to mine ~/.claude/, ~/.codex/, or any other local path. I operate only within this chat session. What I can do is help you build this team from what you tell me about your work. If you want to proceed, here is the plan: **Phase 1 - MINE (requires your input):** You need to provide the raw material. Paste or summarize: - Your recurring tasks and workflows - Tools you use vs. ignore - Corrections you keep giving - Projects that shipped vs. died - Any existing rules or prompt files you want included I will compile this into `team-build/mine-report.md` (as text you can save). **Phase 2-6 - CLASSIFY through HAND OFF:** Once I have your input, I will generate: - `classification.md` - skills/plugins/prompt styles/agents/junk - `design.md` - mission, agent roles, approval gates, kill list - `build-log.md` - role files, skills, plugin specs, routing rules - `verification.md` - test scenario and results - `HANDBOOK.md` - full operating manual **Hard rules I can commit to right now:** 1. Human gates - I will always ask before any send/pay/post/deploy/delete/connect 2. No secrets - I will never print or read .env, tokens, cookies; filenames only 3. Proof of done - I will always provide file path, diff, or verified response 4. Corrections saved - I will track every correction you give me 5. Everything to team-build/ - nothing lives only in chat **To start, please provide:** 1. Your top 5-10 recurring tasks 2. Tools you use daily vs. ones you've abandoned 3. 3-5 corrections you've had to repeat 4. One project that shipped and one that died (why?) Once you give me this, I will build the full team structure.