← ClaudeAtlas

architectlisted

Use when the user invokes `$architect`, usually because they are unclear about what to build or how it should be shaped. Works out the requirement, traces the system and who is affected, and designs before any code. Also surveys existing code for deepening opportunities. Stops with an agreed design for `$full-stop` or `$full-send`.
RIJUDEVx/agent-skills · ★ 1 · AI & Automation · score 77
Install: claude install-skill RIJUDEVx/agent-skills
# Architect Design before implementing. Sketch types, function signatures, class shapes, and module boundaries with `not implemented` bodies and pseudocode. Synthesize across multiple model perspectives, then fill in code against the chosen sketch. If implementation proves the sketch wrong, throw it out and redesign. ## Start Open a todolist with one entry per phase before starting. 1. Ground 2. Sketch 3. Agree 4. Implement 5. Scrap ## Phase A: Ground the problem Build a real mental model of every system the new code touches. Trace the real code: entry points, data flow, state, and who owns what, with `file:line` references. Naming a file isn't grounding. Produce the traced model. If the design redefines ownership or layering, also read the git history and PRs for the existing shape so the rationale becomes a constraint, not a guess. Skip Phase A only when the work is genuinely greenfield with no surrounding system to integrate. Then work out who this touches: use the Blast radius mode of the `interrogate` skill on the planned change. Its findings are constraints on the sketch. ## Phase B: Sketch Start at least two independent sub-agents, on different model families when available, with the design-sketch task and the Phase A findings. Pass `references/runner-prompt.md` as each one's prompt. If you cannot start them, produce the candidates yourself. Each candidate produces a design package shaped per `references/rationale-template.md`. In T3 Code, pick the runners