<!-- adk-managed-skill -->
Creating LWC Components from Design
Orchestrate the end-to-end creation of a new Lightning Web Component from a design input (Figma, PRD, Aura source, or user description). This is the top-level workflow skill — it sequences the specialized sibling skills that own each stage. Org-aware data (LDS schema introspection, design-frame inspection) is resolved by handing off to
experience-lds-data-requirements-generate
or via design-tool URLs the user provides.
When to Use
- Building a brand-new LWC from a Figma frame, PRD, or written spec.
- Migrating an Aura component to LWC as a fresh build (in-place Aura → LWC porting is out of scope for this skill).
- Creating a component whose requirements are still partly implicit and need distillation into a PRD before coding.
Do NOT use this skill for:
- Refactoring an existing LWC (use ).
- Aura → LWC in-place migration (out of scope for this skill).
- Pure styling or data-layer work (use the specialized skills directly).
Prerequisites
- At least one design input: Figma URL, PRD markdown, Aura component source, or a textual spec.
- Target path (module folder) for the new LWC.
- For data-backed components: access to the org so
experience-lds-data-requirements-generate
and experience-lds-best-practices-apply
can resolve the schema and adapter shapes.
Knowledge Bases
- references/prd-analysis-template.md — PRD section skeleton (copy it verbatim when producing a PRD).
- references/figma-to-prd-blueprint.md — Figma-specific guidance for translating a frame into PRD sections (componentName/tagName, contentRequirements, dataRequirements, interactions, componentCommunication, states, accessibility/responsiveness/styling/localization/security). Read this before Phase 1.2 when the input is a Figma design.
Workflow (mandatory five phases)
Phase 1 — Gather PRD & requirements
Goal: produce a consolidated PRD that every later phase consumes.
-
Obtain raw requirements — collect the PRD, design spec, Figma URL, Aura source, or user text.
-
Figma → PRD (if applicable): follow references/figma-to-prd-blueprint.md for the full Figma-frame analysis and PRD section guidelines. Inputs you need from the user: the Figma URL, a screenshot of the target frame, and (if Dev Mode is available) the metadata export for the node. Translate into the PRD skeleton from references/prd-analysis-template.md using the section-by-section guidance in the blueprint.
-
Aura → PRD (if migrating): enumerate the Aura component's functionality that must be preserved — markup, controller/helper actions, events, attributes, and wired data — and feed that inventory into the PRD as explicit requirements. (In-place Aura → LWC porting is out of scope; this step only captures behavior for a fresh build.)
-
Data requirements (if the component reads/writes data): hand off to
experience-lds-data-requirements-generate
. That skill produces a validated data specification (object/field API names, recommended LDS API, implementation approach). Paste its output verbatim into the PRD.
-
Adapter exploration: hand off to
experience-lds-best-practices-apply
for the adapter selection rules (UI API vs GraphQL vs Apex) and the recommended wiring for the chosen approach.
-
Naming:
must be camelCase (e.g.,
) and
must be its kebab-case form (e.g.,
). Validate both with the bundled script — do not eyeball the check:
bash
"<skill_dir>/scripts/check-component-name.sh" <componentName> <tagName>
The script exits nonzero (with an actionable stderr message) if either name is malformed or if the kebab form of
does not equal
.
Deliverable: a comprehensive PRD covering purpose, content, data, interactions, states, a11y, responsiveness, styling direction, localization, and security. Keep it checked into the workspace (e.g.,
packages/skills/<skill>-workspace/<iteration>/PRD.md
).
Phase 2 — Generate component code
Goal: initial
,
,
,
that strictly reflect the PRD.
- Hand off to with the PRD content as the spec. That skill owns PRD → code translation (events, getters, , , AI metadata) and is the authoring source of truth.
- Use the SLDS decision hierarchy from
design-systems-slds-apply
:
- Prefer a matching Lightning Base Component (
experience-lwc-base-components-integrate
).
- Otherwise pick an SLDS Blueprint or utility class (per
design-systems-slds-apply
).
- Otherwise write custom CSS using SLDS styling hooks (also covered in
design-systems-slds-apply
).
- Re-confirm the authoring baseline by walking the checklist before proceeding.
- Translate every PRD data requirement to either a wire adapter (UIAPI / GraphQL) or an explicit TODO.
- Adhere strictly to the PRD:
- Only include features and behaviors the PRD describes.
- Mark uncertain areas with comments that quote the PRD language raising the ambiguity.
- Add extensive code comments explaining intent.
- Do not split the component into more sub-components than the PRD implies.
Deliverable: a first-pass LWC bundle in the target path.
Phase 3 — Optimize component
Goal: apply performance, maintainability, and best-practice fixes.
- Walk the optimization checklist below before any other review:
- Single responsibility, encapsulation, reusability.
- Minimize DOM operations; batch updates via properties.
- Audit event handlers and lifecycle hook usage ( guards).
- Consider lazy loading for heavy children.
- Hand off to for the LWC best-practices review pass (anti-patterns, reactivity, composition).
- Hand off to the compliance suite for the pre-ship review —
experience-lwc-accessibility-validate
(a11y), experience-lwc-security-validate
(LWS + Product Security), and experience-lwc-rtl-validate
(RTL i18n). Run them together for the full quality pass.
- If the component touches data, hand off to
experience-lds-best-practices-apply
for cache/consistency and referential-integrity checks.
- Apply every accepted finding. Keep the PRD as the source of truth — do not add scope under the guise of optimization.
Phase 4 — Lint, format, compile-check
Condition: only run steps whose tooling is configured in the project.
-
Detect project tooling — invoke the bundled detection script from the project root and read its
lines. Do NOT eyeball
/ dotfiles in prose:
bash
"<skill_dir>/scripts/detect-project-tools.sh" <projectRoot>
Sample output:
text
eslint=yes
prettier=yes
cursor-rules=no
lwc-compiler=yes
Run the substeps below only for tools reported
; skip the rest.
-
ESLint (if
) — run and fix all violations.
-
Prettier (if
) — run for consistent formatting.
-
Cursor rules (if
) — apply every rule the project ships.
-
LWC compiler (if
) — run via local dev server or SFDX; resolve every syntax/template error before moving on.
Deliverable: clean, validated component code.
Phase 5 — Create tests
Hand off to
experience-lwc-accessibility-jest-run
for automated accessibility Jest coverage; add general-purpose Jest coverage in the same pass following the
test guidance, plus UTAM page object generation if the team requires it.
Deliverable: the LWC bundle with a passing test suite at or above the project's coverage threshold.
Definition of Done
A new component built from this workflow is "done" only when every item below is true. Treat this as the canonical readiness checklist — copy it into the PR description so reviewers can confirm each line.
Code completion — no implementation gaps:
- Every PRD requirement is either implemented or explicitly annotated with a and a linked tracking item. No silent gaps.
- No , , or / / left in production paths.
- No commented-out code blocks, no empty function bodies, no placeholder values, no dummy data, no unreferenced imports.
- No regions or third-party-library escape hatches without a comment explaining why a native LWC pattern wasn't used.
Compliance and quality:
- AI metadata passes the audit in (component-wide set, plus an
<ai><property name="…" aiDescription="…"/></ai>
entry for every member exposed through ; no marketing language).
- SLDS styling passes
design-systems-slds-apply
verification — no raw hex / px values, only styling hooks and SLDS utility classes.
- Accessibility pass complete:
experience-lwc-accessibility-validate
for source review + experience-lwc-accessibility-jest-run
for automated tests, both green.
- Security + RTL pass complete:
experience-lwc-security-validate
+ experience-lwc-rtl-validate
, both green.
Data + tests:
- Data layer verified against the referenced LDS adapters (every wire / imperative call is documented in the PRD's data section and matches one of the adapters from
experience-lds-best-practices-apply
or experience-lds-data-requirements-generate
).
- Jest tests green at the required coverage level for the project. New tests cover every surface, every dispatched event, and every error path (failed wire, failed apex, validation rejection).
- UTAM page objects produced for any UI flow that needs cross-component browser-level testing (skip if not needed).
Cross-References
- Skills chained by this workflow (in phase order):
experience-lds-data-requirements-generate
— Phase 1.4 data spec (and Phase 1.5 adapter exploration alongside experience-lds-best-practices-apply
).
design-systems-slds-apply
, experience-lwc-base-components-integrate
— Phase 2 styling decisions.
- — Phase 2 authoring baseline + Phase 3 best-practices review + Phase 5 AI-metadata audit.
experience-lds-best-practices-apply
— Phase 2/3 data-layer adapter selection and consistency review.
experience-lwc-accessibility-validate
, experience-lwc-security-validate
, experience-lwc-rtl-validate
— Phase 3 a11y/security/RTL review.
experience-lwc-accessibility-jest-run
— Phase 5 automated a11y test generation.
- Optional: add o11y instrumentation as a separate pass once the component stabilizes.
experience-lwc-typescript-migrate
— optional once JS is green.
- When the new component ships behind a flag, gate it with a feature flag during rollout.
experience-lwc-api-docs-generate
— once the public API surface is stable.
- Org-aware inputs used by this workflow:
- Figma URL + screenshot (and, if available, the developer's Dev Mode metadata export) — Phase 1.2 Figma input.
experience-lds-data-requirements-generate
owns the org-schema introspection and data-spec validation used in Phase 1 when the component needs org-backed data — hand off to that skill rather than duplicating its work here.
Examples
Phase 1 PRD skeleton (fill from the Figma/PRD/Aura input)
text
Component: productCard (<product-card>)
Purpose: Render a compact summary of a product with quick actions.
Content requirements
- Hero image (Product.HeroImage__c)
- Title (Product.Name)
- Subtitle (Product.Tagline__c)
- Primary CTA button ("Add to cart")
- Secondary CTA icon button ("Favorite")
Data requirements
- Input: @api recordId (Product Id)
- Adapter: getRecord (UIAPI) with fields Name, Tagline__c, HeroImage__c
- Events out: addtocart{detail.recordId}, favorite{detail.recordId, detail.value}
States
- Loading (data not yet resolved)
- Error (adapter error)
- Empty (no record found)
- Default
Accessibility
- Title uses <h3>
- Icon-only button carries aria-label="Favorite"
- Card is a labelled region (role="group", aria-labelledby)
Responsiveness
- Full width < 480px
- Side-by-side image + text >= 480px
Styling
- Uses lightning-card wrapper
- Surface color: --slds-g-color-surface-container-1
- Shadow: --slds-g-shadow-1
Phase 2 skeleton
javascript
import { LightningElement, api, wire } from 'lwc';
import { getRecord } from 'lightning/uiRecordApi';
import NAME from '@salesforce/schema/Product__c.Name';
import TAGLINE from '@salesforce/schema/Product__c.Tagline__c';
import HERO from '@salesforce/schema/Product__c.HeroImage__c';
const FIELDS = [NAME, TAGLINE, HERO];
export default class ProductCard extends LightningElement {
@api recordId;
@wire(getRecord, { recordId: '$recordId', fields: FIELDS })
record;
get hasRecord() { return this.record?.data != null; }
get isLoading() { return !this.record; }
get hasError() { return !!this.record?.error; }
get name() { return this.record?.data?.fields?.Name?.value ?? ''; }
get tagline() { return this.record?.data?.fields?.Tagline__c?.value ?? ''; }
get heroUrl() { return this.record?.data?.fields?.HeroImage__c?.value ?? ''; }
handleAddToCart() {
this.dispatchEvent(new CustomEvent('addtocart', { detail: { recordId: this.recordId }, bubbles: true, composed: true }));
}
handleFavorite(event) {
this.dispatchEvent(new CustomEvent('favorite', { detail: { recordId: this.recordId, value: event.detail.value }, bubbles: true, composed: true }));
}
}
Verification
- Every PRD section is traceable to at least one block of code or an explicit TODO.
- Phases 1–5 were executed in order; no step was skipped.
- Accessibility, SLDS, data-layer, and AI metadata reviews all pass.
- Coverage meets or exceeds the project's threshold.
- The component renders correctly in
experience-lwc-runtime-observe
across the responsive breakpoints enumerated in the PRD.