← ClaudeAtlas

architect-buildlisted

Manual-only skill. Use only when the user explicitly invokes the architect-build skill to execute one sealed plan in order, preserve its hard functional boundary, adapt cautiously when the code impact scope expands, record execution-time decisions truthfully, and request a user decision only when the functional boundary cannot be preserved. Do not auto-trigger from a generic implementation request.
vortezwohl/Architect · ★ 27 · Web & Frontend · score 72
Install: claude install-skill vortezwohl/Architect
# Architect Build Use this skill only as the manually selected `architect-build` stage. Its job is to execute the sealed plan from start to finish, preserve its hard functional boundary, and keep execution state and logs truthfully aligned with actual progress. It may make cautious, local, execution-time minimal design decisions when the recorded code impact scope proves incomplete, but it does not reopen earlier stages or silently change the approved functionality. It is stage 3 and the final stage of the one-way flow `architect-design -> architect-propose -> architect-build`. ## Defined Terms - `functional boundary`: the approved target functionality, protected related functionality, explicit non-goals, compatibility obligations, and hard-stop condition. It is a hard execution constraint. - `code impact scope`: the recorded expected paths, symbols, configuration, tests, and callers likely to be affected. It is a coverage-oriented reference, not a permission list or a prohibition list. - `impact-scope adaptation`: a cautious execution-time assessment and minimal implementation decision made when required work extends beyond the recorded code impact scope but remains inside the functional boundary. - `functional-boundary exception`: a user-approved runtime decision made only after build proves that the target cannot be completed without changing protected functionality or a non-goal. ## Manual Invocation Only - Run this skill only when the user explicitly