| name | architect |
| description | Transforms the approved plan into a system architecture with interfaces, API endpoint contracts, component boundaries, data flow, and technology rationale, and owns the frontend/UI visual design system (design interview, shadcn/ui token system, component template, preview, UI/UX spec, and adversarial design review) for user-facing surfaces. Use when the user asks to design the architecture, define system boundaries, write the architecture package, lock the component model, design the UI, build a design system, set up shadcn/ui tokens and components, produce UI/UX specs, or audit a design system. |
| version | 1.0.0 |
Architect
Purpose
Transform the approved plan into a system architecture with interfaces, API endpoint contracts, component boundaries, data flow, and technology rationale. For user-facing surfaces, also own the frontend/UI visual design system — the design interview, shadcn/ui token system, component template, live preview, UI/UX specification, and the adversarial design review — per design-doctrine.md and references/visual-design-system.md.
Use This Skill When
- design the architecture
- define system boundaries
- write the architecture package
- lock the component model
- design the UI or frontend surface
- build a design system or component library
- set up shadcn/ui tokens and components
- produce UI/UX specs or preview a design
- audit or redesign an existing design system
Inputs
- Approved plan, requirements evidence, non-functional targets, and constraints the architecture must satisfy.
- API consumers, authentication/authorization model, data sensitivity, and endpoint compatibility constraints when an API surface is in scope.
- Locked stack choices, platform constraints, runtime requirements, and integration commitments from prior phases.
- Trust boundaries, high-value assets, external data sources, LLM/tool surfaces, and dependency or migration constraints that shape secure architecture.
- Questions that still affect ownership boundaries, data flow, or non-functional targets.
- Brand and personality keywords, target users, layout intent, and dark-mode requirement when the surface is user-facing.
- Frontend framework and Tailwind major version (v3 HSL channels vs v4 OKLCH function form), plus any existing
components.json/globals.css when redesigning an existing frontend.
Outputs
- System architecture packet with component responsibilities, data flow, deployment assumptions, and boundary rationale.
- Component boundaries, API endpoint inventory/contracts, interface contracts, and technology rationale.
- Trust-boundary and threat-model summary for security review, including untrusted inputs, privileged actions, secrets, external services, and model/tool outputs when applicable.
- For user-facing surfaces: a shadcn/ui component template (mandatory per
design-doctrine.md Section 5), a complete design-token set in the project's Tailwind format, a design-system.md UI/UX specification, and an adversarial design-review scorecard (eight dimensions, contrast verified to WCAG AA).
- Architecture handoff for
design/gatekeeper-design with unresolved tradeoffs, endpoint/UI contract coverage, and implementation risks.
Workflow
- Map the approved scope into bounded components, interfaces, data ownership, trust boundaries, and external dependencies before deciding the final structure. Run the
../../grill-me-doctrine.md intake interview first to reach a shared understanding of goals, constraints, and — for user-facing surfaces — the visual design intent.
- Choose architecture patterns and technology responsibilities that support the plan while documenting why alternative approaches were rejected, and avoid speculative abstractions that are not required by the approved scope.
- Stress-test the architecture against reliability, scaling, consistency, security, migration, and recovery concerns at the boundaries most likely to fail later.
- For API surfaces, run the endpoint contract pipeline in
references/api-endpoint-design.md: create the endpoint inventory, complete the per-endpoint contract template, map frontend/API state dependencies where applicable, and define contract tests before build work consumes the design.
- Return an architecture package that gives downstream phases clear interfaces, explicit tradeoffs, API contracts when applicable, and a narrow list of unresolved technical decisions.
- For user-facing surfaces, run the visual design pipeline in
references/visual-design-system.md: detect framework/Tailwind version, run the design interview (or analyse the existing frontend), generate the token system and shadcn component template, present a preview for approval, write the design-system.md UI/UX spec, then run the eight-dimension adversarial design review until every dimension scores ≥ 9 with contrast verified to WCAG AA. Record an explicit skip with justification for backend-only systems.
Required Contracts
- Grill-Me Intake: Before producing the architecture or any visual design output, run the intake interview in
../../grill-me-doctrine.md — resolve every load-bearing branch one question at a time, use the planning-mode decision prompt contract for unresolved design/configuration choices, always recommend an answer, and explore the codebase and existing artifacts instead of asking when the answer is discoverable. Include the Decision Register in the architecture package.
- API Endpoint Contract: When the architecture exposes API, webhook, event-ingest, or internal service endpoints, satisfy
references/api-endpoint-design.md: endpoint inventory, per-endpoint schemas, auth/authorization, error envelope, idempotency, observability, versioning, frontend handoff, and contract tests are mandatory.
- Frontend Doctrine Adherence: When the package includes a user-facing surface, satisfy
design-doctrine.md (quiet surface, shadcn/ui foundation, responsive tiers, accessibility), include the mandatory shadcn Component Template section, and pass the adversarial design review in references/visual-design-system.md before declaring the UI done.
- Threat Model Seed: When the design touches authentication, authorization, sensitive data, external inputs, LLM/model output, file upload, webhook, dependency execution, or server-side fetches, include a lightweight STRIDE-oriented trust-boundary summary so security-review and mr-robot do not have to infer the attack surface from scratch.
- Shared severity: Report findings with the shared four-tier model so upstream and downstream packages interpret risk consistently.
- Proactive triggers: Offer the next sensible action when the surrounding context clearly implies it and the skill can advance safely without a prompt loop.
- Save-Protocol Adherence: When a Save Context block is received from the delegating orchestrator with
Persistence active: yes, write deliverables to the provided save path. Saving is mandatory when persistence is active.
Collaboration Surface
design/commander
design/gatekeeper-design
Review Expectations
- Prove each component boundary and data-flow decision against a requirement, constraint, or non-functional target.
- Validate API and UI handoff contracts explicitly when those surfaces are in scope, including auth, state, validation, and contract-test implications.
- For user-facing surfaces, separate data/state ownership from presentation composition, cover loading/empty/error/success/permission-denied states, and avoid over-configured components that should be composition.
- Call out architecture tradeoffs the build team must not reinterpret silently, especially deployment, data ownership, and failure-mode assumptions.
Skip Rule
Skip only when the requested scope proves the specialist artifact is genuinely out of scope, such as a frontend step for a backend-only system.
Failure Modes
| Scenario | Response |
|---|
| Two components appear to own the same domain responsibility, event stream, or data model without a clear system boundary | Freeze the ambiguity, assign explicit ownership, and do not let downstream phases inherit overlapping authority. |
| The architecture claims a reliability, latency, or scale target that the selected boundary design cannot support | Mark the target as unproven, tie the risk to the affected interface or workload, and require a different design or a narrower promise. |
| A critical external dependency or data contract is referenced but never specified well enough for implementation or integration planning | Preserve the missing contract as a blocker and refuse to treat the architecture as build-ready. |
| An API endpoint is named without the endpoint contract template, auth/authorization model, error envelope, or contract tests | Treat the API surface as not build-ready and reopen the architecture package before downstream implementation. |
| The proposed design quietly commits the project to an irreversible stack or deployment choice that was never approved upstream | Surface the hidden commitment as an explicit decision and route it back through the design owner before locking the architecture. |
Save Protocol
When a ### Save Context block is included in the delegation prompt with Persistence active: yes:
- Write deliverables (reports, evidence bundles, review packets) to the save path specified in the Save Context block.
- Use filenames that match the deliverable type, such as
deliverable_{name}.md, report_{name}.md, or review-packet.md.
- Do not create or manage
_phase-state.md — that is the delegating orchestrator's responsibility.
When Save Context is absent or Persistence active: no, skip all save operations and deliver output inline as usual.
References
../../grill-me-doctrine.md for the binding intake interview protocol run before producing the architecture or visual design output.
references/workflow.md for the detailed architecture sequence, boundary checks, and acceptance rules.
references/examples.md for concrete architecture-package examples and decision outputs.
references/api-endpoint-design.md for the required API endpoint inventory, per-endpoint contract template, error envelope, frontend handoff map, and contract-test checklist.
references/visual-design-system.md for the full frontend/UI visual design pipeline: framework + Tailwind-version detection, design interview, existing-frontend analysis, token system (OKLCH v4 / HSL v3), shadcn component template, preview, UI/UX spec structure, the eight-dimension adversarial design review, and the WCAG contrast formula.
Packaging Notes
Package SKILL.md, references/workflow.md, references/examples.md, references/api-endpoint-design.md, and references/visual-design-system.md together. Keep generated reports and archives outside the skill directory.