← All skills

plan-ui-change

>

Class skill ID skill/plan-ui-change Version 1.0.0 Lifecycle stable Author anthropics
/plan-ui-change
codedotnetengineering dotnet-blazorcsharpmicrosoft
System prompt fragment

Plan a Blazor UI Change

When asked to build a complex UI feature, plan the component decomposition first, then immediately implement it. A single monolithic page component is almost never the right answer — break the UI into focused, composable components.

Planning Workflow

Step 1 — Map the Visual Regions

Read the request and identify every distinct visual region. Each region that has its own data, behavior, or layout responsibility is a candidate component.

Draw the component tree:

InventoryDashboard          (page — owns data, orchestrates layout)
├── StockSummaryBar         (read-only stats: total items, low-stock count, value)
├── InventoryFilters        (search box, category dropdown, stock-level toggle)
├── InventoryTable          (sortable table of products)
│   └── InventoryRow        (single product row with inline edit/delete)
└── AddProductForm          (slide-out form for new products)

Rules for identifying components:

  • Distinct responsibility — a region owns its own state or behavior → separate component
  • Repeated structure — items in a list, cards in a grid → extract the item template
  • Independent interactivity — a section that handles user input separately from its siblings → separate component
  • Size — any section that would exceed ~150 lines of markup on its own → split it

Step 2 — Classify Each Component

For every component in the tree, determine:

Component Action Render Mode State Owned Lines (est.)
InventoryDashboard Create InteractiveServer product list, filter state ~80
StockSummaryBar Create (inherits) none — receives data ~30
InventoryFilters Create (inherits) search text, selected category ~60
InventoryTable Create (inherits) sort column, sort direction ~50
InventoryRow Create (inherits) inline-edit mode flag ~60
AddProductForm Create (inherits) form model ~80

A page component that exceeds ~200 lines of combined markup + code is too large. If your estimate puts a single component above that, split further.

Step 3 — Design Data Flow

Identify the state owner for each piece of data, then map how it flows:

InventoryDashboard (owns: products[], filters)
  │
  ├─ [Parameter] products ──→ StockSummaryBar (reads aggregate stats)
  │
  ├─ [Parameter] filters ──→ InventoryFilters
  │   └─ EventCallback<Filters> OnFiltersChanged ──→ InventoryDashboard
  │
  ├─ [Parameter] filteredProducts ──→ InventoryTable
  │   └─ [Parameter] product ──→ InventoryRow
  │       ├─ EventCallback<Product> OnSave ──→ InventoryTable ──→ InventoryDashboard
  │       └─ EventCallback<Product> OnDelete ──→ InventoryTable ──→ InventoryDashboard
  │
  └─ EventCallback<Product> OnProductAdded ←── AddProductForm

Rules:

  • Data always flows down through [Parameter]
  • Events always flow up through EventCallback<T>
  • The page/parent owns the data and passes filtered/transformed views to children
  • Children never mutate parameters — they notify the parent via callbacks
  • If data must cross more than 2 levels without intermediate components needing it, use a cascading value or a scoped service

Step 4 — Identify Reuse Opportunities

Before creating a new component, check if an existing component in the project can serve the purpose. Look for:

  • Existing list-item components that match the structure
  • Shared filter/search components already in the project
  • Generic components (e.g., DataTable<T>, Pagination) that accept templates

If a component will be used in more than one page, place it in a Shared/ or Components/ folder.

Step 5 — Order the Implementation

Build bottom-up — leaf components first, then parents that compose them:

  1. Models/DTOs — define the data shapes
  2. Services — data access, business logic (interface + implementation)
  3. Leaf components — components with no children (InventoryRow, StockSummaryBar)
  4. Container components — components that compose leaves (InventoryTable, InventoryFilters)
  5. Page component — wires everything together, registers routes
  6. Configuration — DI registration, render mode setup

Each component should be independently compilable. Never reference a component that doesn't exist yet.

Output Format

Present the plan briefly, then immediately proceed to implement — never stop at just the plan or ask for confirmation before writing code. The plan is a thinking tool, not a deliverable.

## Component Plan: [Feature Name]

### Component Tree
[ASCII tree showing parent-child relationships]

### Component Table
| Component | Action | Render Mode | Purpose | Est. Lines |
|-----------|--------|-------------|---------|------------|
| ... | ... | ... | ... | ... |

### Data Flow
[State owner] → [Parameters down] → [EventCallbacks up]

### Implementation Order
1. [First file to create — why]
2. [Second file — why]
...

After outputting the plan, immediately begin implementing the components in the order listed. Do not wait for approval or ask "shall I proceed?" — the plan is a guide for you to follow, not a proposal for the user to approve.

Anti-Patterns to Avoid

Anti-Pattern Why It's Wrong Correct Approach
One page component with 500+ lines Impossible to test, reuse, or maintain Decompose into focused components
Passing 10+ parameters through intermediate components Parameter drilling obscures intent Use cascading values or a scoped state service
Child component fetching its own data from an API Multiple components making redundant calls Parent owns data, passes via parameters
Inline rendering of list items with complex markup Duplicated logic, no reuse, hard to test Extract item template into its own component
Building everything in one file then "refactoring later" Refactoring rarely happens; the monolith ships Plan the decomposition upfront
Generic components for one-off usage Over-engineering adds complexity Only extract generics when reuse is proven

Guidelines

  • Plan briefly, then implement. Write a concise component table and data flow map, then immediately create the .razor files — never stop at just the plan.
  • Prefer many small components over one large one. A component with a single clear purpose is easier to understand, test, and reuse.
  • State ownership is the first decision. Before writing fetch logic, decide which component owns the data.
  • Build bottom-up. Create leaf components first so parent components can reference them immediately.
  • Name components after what they render, not what they do internally: ProductCard not ProductRenderer, OrderFilters not FilterHandler.
Attribution: anthropics. View original source ↗. License: Apache-2.0 or as stated by the originating repository.