lamina

Use only when explicitly invoked as lamina-design. Turn an incomplete product idea or brownfield change into a minimum sufficient product behavior graph: actors, entities, operations, workflows, rules, dependencies, decisions, persona perspectives, and distinct risks; then generate an implementation-ready contract.

Use only when explicitly invoked as lamina-init. Turn an incomplete product idea into usable business context and evidence-grounded personas, asking only high-leverage questions and labeling provisional assumptions for later product-graph design.

Product design workflows — domain contracts, implement brief, verify loop. Load via Read or Skill tool when /lamina-* workflows need it.

Use only when explicitly invoked as lamina. Route product design for developers building with AI — design, verify, or direct capability answers.

Use only when explicitly invoked as lamina-verify. Verify a live or brownfield product against its product graph: critical promises, reachable workflows, authority, invariants, state integrity, recovery, accessibility, and contract drift; emit evidence-backed fixes before merge.

Business context UX guidance. Use when bootstrapping Lamina for a project; answering business questions UX work needs; updating context after a pivot or scope change.

Lamina Problem Router — product design at the intersection of UX, product rules, and systems thinking.

Map prerequisite, data, lifecycle, and reachability dependencies between typed product-graph nodes, including explicit behavior when a dependency is unmet.

Form behavior in contracts — validation timing, field semantics, and recovery UX. Use when data entry blocks workflows or causes scenario gaps.

Consistency across actor views — student, admin, and system surfaces must reflect the same domain truth. Use in multi-actor products when roles see different slices of one system.

Downstream effects of state changes — notifications, updates to related entities, and cross-actor handoffs. Use when one operation must trigger updates beyond the primary screen.

Define trustworthy date, time, deadline, expiry, recurrence, and timezone behavior. Use when a product accepts local times, crosses actor timezones, schedules recurring work, or triggers behavior from a clock.

Trust signals in contracts — transparency for payments, irreversible actions, and sensitive data. Use when actors hesitate on verify walks or scenarios lack honesty UX.

Specify and verify accessibility behavior on product-graph surfaces and workflows, including semantics, keyboard reachability, focus, errors, async status, and non-visual state communication.

Copy and labels in contracts — scan-first labels, error text, and empty-state messaging. Use when screen copy blocks actor walks.

Actions and reversibility in contracts — primary/destructive actions, undo policy, confirmations. Use when destructive ops lack scenarios.

Prioritize contract gaps and verify findings — impact × effort, primary actor filter. Use when reconciling conflicts or ranking findings[].

Agent-native design loop — domain contract, external build, verify, iterate. Not human workshop ceremony.

Signifiers and affordances in screen contracts — actors must see what they can do. Use when actor walks report execution/evaluation gulfs.

Derive distinct product-risk scenarios from operations, dependencies, authority, lifecycles, concurrency, destructive actions, and recovery without generating exhaustive or duplicate edge-case lists.

Error and recovery UX in scenarios — slips vs mistakes, root cause in design not actors. Use when mapping failure scenarios in run.json.

Capability discovery — map requests to domain operations, workflows, and actors. Not generative research methods for human labs.

Live product grounding — repo and browser observation of real behavior. Not ethnography logistics.

Design reachable actor workflows from typed product operations, dependencies, and terminal outcomes, including complete cross-actor handoffs and recovery.

Expert lens review — parallel contract, a11y, and invariant checks. Not Nielsen heuristic checklist theater.

Entity organization and retrieval — how actors find domain objects in screens and nav. Use when IA mirrors files or tables instead of tasks.

Walkthrough evidence — structured capture from browser/repo sessions. Not human session notes.

Product invariants — rules that must always hold, impossible states prevented, errors defined out of existence. Use when defining what can never happen in the product.

High-impact intervention points in product design — rules, information flows, and goals over UI tweaks. Use when local fixes fail to change behavior.

Domain and feature boundaries — hide complexity behind clear ownership, pull complexity away from users. Use when splitting entities, actors, and responsibilities.

Wayfinding in contracts — persistent nav, orientation, and hierarchy per screen. Use when actors get lost in verify walks.

First-run and learnability — primary path without training, progressive power features. Use when new actors can't complete first workflow in verify.

Product density — how much complexity the domain tolerates on each surface. Not sovereign/transient platform essays.

Design intake — align user request to domain scope, actors, and success outcome. Not pre-fieldwork research questions.

Represented model matches domain — UI must not imply illegal states or permissions. Use when run.json domain and screens diverge from implementation shape.

Progressive disclosure in contracts — essential vs advanced actions per actor. Use when novices overwhelmed or experts blocked on power features.

Translate product intent into traceable critical promises, graph nodes, observable scenarios, assumptions, and scope without prescribing unnecessary implementation details.

Simulation planning — plan actor walks, invariant probes, and walkthrough capture for verify. Not recruitment logistics.

Evidence scoping — what grounding is needed (repo, walkthrough, user input) vs assumption. Not human study scope.

Simulation synthesis — merge parallel actor-walk and walkthrough results into findings[]. Not affinity mapping human interviews.