plan-implementationlisted

# Plan the implementation Use this skill to **break** a goal into an implementation plan — add the work in small, controlled increments toward the outcome, each one producing an observable signal that it landed. The output is an **ordered task plan**: the increments, their dependencies and risk, and the acceptance check that tells you each is done. Governing intuition: **each increment must produce a signal you can see flip.** A task whose "done" you cannot observe is not verifiable — reslice it until it has an acceptance check. Prefer the smallest slice that proves something real and leaves the system working over a big step you cannot verify midway; and order by what unblocks or de-risks the rest, not by what is easy. **Input**: the goal to plan toward — a feature, fix, or change — plus its hard constraints; if none is given, ask what outcome the plan should reach. **Boundary**: plans only — writes no code, scaffolds nothing, runs no build; produces the plan an implementer executes. ## Workflow ### Phase 0: Frame the goal and constraints - State the goal as an **outcome** — what will be true when it is done — not as a task list. - Capture hard constraints: behaviour that must not break, interfaces to honour, deadlines, and the tech the team already owns. - Check the upstream boundary: unfamiliar code the plan touches → `map-codebase` first; an unsettled component structure → `design-boundaries` first; a behaviour-preserving decomposition → `plan-split`; a data m