What problem does it solve? Long AI chat sessions turn into unstructured streams of text where requests get lost, unanswered questions go unnoticed, and neither side can reference what was said earlier. This Skill turns every prompt and response into addressable, numbered thoughts with explicit classes, and defines a shared work folder with clear ownership rules. ## Core Features & Use Cases - Addressable thoughts: Every prompt is parsed into numbered thoughts (P⟨n⟩.M) and every agent reply into numbered response thoughts (R⟨n⟩.M), each tagged with a class emoji indicating what kind of response is expected. - Turn discipline via task lists: Each turn opens with a fixed three-item todo list (analyze prompt, do work, write response) that is rebuilt as work is discovered, keeping the plan visible before execution starts. - Work folder conventions: Defines INDEX.md (maintained by the agent as a work map replacing plan.md) and NOTES.md (maintained by the human for personal anchors), with copy-ready templates. - Use Case: During a multi-session project, a user references R7.2 to dispute a specific agent claim instead of re-explaining context, and the agent's prompt analysis surfaces a misunderstood requirement before any files are touched. ## Quick Start Adopt the duet-work-full conversation format: parse my next prompt into numbered classified thoughts before doing any work, and reply with numbered response thoughts.
