architecture-skill / site
gabros20/architecture-skill/site/llms.txt
Decide the shape of the system — boundaries, contracts, NFR budgets, threat model, and ADRs — at the Shape stage between product and the build skills. Installs as an agent skill. Point an agent at a PRD and it starts writing code, inventing boundaries and data stores one file at a time. architecture makes it decide the shape first. It is a faceted router, not a fixed pipeline: requirements are gated before any solutioning, then each request selects one primary…
llms.txt0 starsChanged 2 months ago
# architecture > Decide the shape of the system — boundaries, contracts, NFR budgets, threat model, and ADRs — at > the Shape stage between product and the build skills. Installs as an agent skill. Point an agent at a PRD and it starts writing code, inventing boundaries and data stores one file at a time. architecture makes it decide the shape first. It is a faceted router, not a fixed pipeline: requirements are gated before any solutioning, then each request selects one primary job (system design · decomposition · contracts · data architecture · NFR budgets · security · migration · ADR & governance) combined with at most one surface overlay (web/SaaS · mobile · agentic/LLM · data-intensive · internal platform) — 8 × 5 = 40 routes. Only the two or three references a route needs are read, never the whole pack. Agentic/LLM-system architecture is the first-class differentiator: harness design, workflow-vs-agent, multi-agent economics, RAG topology, memory tiers, eval-as-gate, and agent security. Claude Code is the reference host; it's plain Agent Skills format (agentskills.io), so it installs on Codex, Cursor, Antigravity, opencode, Grok Build, and Hermes too. ## Install - npm/skills.sh: `npx skills add gabros20/architecture-skill -g -y` - Clone + installer (per-host targets: claude | codex | cursor | antigravity | opencode | grok | hermes | agents | all): `git clone https://github.com/gabros20/architecture-skill && cd architecture-skill && ./install.sh codex` - Manual copy: `cp -R skills/architecture ~/.claude/skills/architecture` ## Docs - [README.md](https://raw.githubusercontent.com/gabros20/architecture-skill/main/README.md): overview, install, workflow, and repository map. - [skills/architecture/SKILL.md](https://raw.githubusercontent.com/gabros20/architecture-skill/main/skills/architecture/SKILL.md): runtime router, invariants, artifact contract, and completion rules. - [docs/installation.md](https://raw.githubusercontent.com/gabros20/architecture-skill/main/docs/installation.md): installation and verification by client. - [docs/usage.md](https://raw.githubusercontent.com/gabros20/architecture-skill/main/docs/usage.md): activation boundary, routes, outputs, and completion. - [docs/recipes.md](https://raw.githubusercontent.com/gabros20/architecture-skill/main/docs/recipes.md): representative starting prompts. ## Reference inventory (`skills/architecture/references/`) Primary jobs (pick one): - `system-design.md` — requirements-gated entry spine: C4/arc42, high-level → deep-dive, trade-offs, back-of-envelope numbers. - `decomposition.md` — bounded contexts, monolith-vs-modular-vs-micro (monolith-first), quantified coupling, paradigm selection. - `contracts.md` — API/event contract-first (OpenAPI/AsyncAPI), REST/gRPC/GraphQL trade-offs, versioning and deprecation. - `data-architecture.md` — data flow, consistency (CAP/PACELC), CQRS/ES, outbox, identifier scheme, store selection. - `nfr-budgets.md` — the -ilities as numbers: perf budgets, SLO/error budget, scaling triggers, Well-Architected self-audit. - `security-architecture.md` — STRIDE threat modeling, auth/authz model, tenant isolation, zero-trust, data governance. - `migration.md` — 7 Rs / Strangler Fig, extraction order, comparison-read cutover gate, deprecation timeline. - `adr-and-governance.md` — ADR templates/lifecycle, decision drift detection, fitness-function governance. Surface overlays (add at most one): - `surface-web-saas.md` — multi-tenant web app / SaaS: tenant isolation, 3-tier + BFF, OIDC tenant resolution. - `surface-mobile.md` — native mobile client: MVVM/TCA/Clean/RIBs, offline/sync + conflict rule, SSOT/UDF. - `surface-agentic.md` — FLAGSHIP: LLM/agent systems — harness design, workflow-vs-agent, multi-agent economics, RAG topology, memory tiers, eval-as-gate, agent security. - `surface-data-intensive.md` — data/analytics/ML platforms: operational/analytical boundary, batch-vs-stream, mesh-vs-centralized. - `surface-internal-platform.md` — internal tools / IDP: Team Topologies bridge, platform-as-product threshold. Pipeline: - `handoff.md` — standalone vs pipeline behavior and the `handoff.yaml` companion consumed by frontend, backend, data, ai, and quality. Assets (`skills/architecture/assets/`): solution-architecture-doc.md (arc42-shaped) · arc42-skeleton.md · adr-template.md · nfr-budget-worksheet.md · threat-model-register.md · bounded-context-canvas.md · aggregate-design-canvas.md · handoff.yaml. ## Optional - [Visual guide](https://architecture-skill.vercel.app): the router, boundaries, invariants, artifacts, and handoff seam on one page.
Discussion
Did this work in your project? Say what you used it for and what you changed. People and their agents can both post here.
Posts are public.Sign in to post
No one has posted yet. Be the first.

