| name | experience-lds-best-practices-apply |
| description | Use when reviewing or implementing Lightning Data Service best practices in an LWC (.js, .html, .js-meta.xml) — UIAPI vs Apex, refreshApex / notifyRecordUpdateAvailable, @salesforce/schema imports, LDS record-form data patterns. TRIGGER on "apply LDS best practices to this LWC", "review this LWC for LDS best-practice issues", "review this component for Lightning Data Service issues", "UIAPI or Apex for this data?", "fix stale data after record save", "sync LDS cache", "use @salesforce/schema for field names", "choose between getRecord and Apex". DO NOT TRIGGER when building a new LWC (use experience-lwc-generate), applying SLDS design tokens (use design-systems-slds-apply), picking or wiring a `lightning-*` base component's props/events/slots generically (use experience-lwc-base-components-integrate — this skill covers only the LDS data-layer rationale, even when the fix involves a base record form), or for security / RTL / accessibility reviews (separate passes). |
| metadata | {"version":"1.0","relatedSkills":["design-systems-slds-apply","experience-lwc-base-components-integrate","experience-lwc-generate"]} |
Applying LDS Best Practices
Apply the Lightning Data Service guidelines to a Lightning Web Component. Three pillars: data consistency, referential integrity, and UIAPI vs Apex. Focused on the UI API path — GraphQL and upstream data-requirements analysis are handled out-of-band.
When to Use
- Reviewing a component's data layer for LDS compliance (hand-rolled forms, stringly-typed field names, un-synchronized Apex + LDS, Apex overuse).
- Implementing CRUD on standard or custom objects.
- Deciding between
getRecord, getRecords, createRecord, updateRecord, deleteRecord, base record form components, or Apex.
- Fixing stale-data bugs after record mutation.
- Adding schema imports (
@salesforce/schema/...) to replace hard-coded field/object names.
Do NOT use this skill for:
- GraphQL query/mutation generation — handled out-of-band today.
- Upstream data-requirements discovery — handled out-of-band today.
- SLDS class / design-token work (use
design-systems-slds-apply).
- Accessibility, security, or RTL review — those are separate passes run with their own tooling.
Prerequisites
- Component path.
- Understanding of the component's data operations (read / write / both) and whether Apex is already involved.
- Access to the org's schema for
@salesforce/schema imports (Setup → Object Manager → <Object> → Details → API Name; or, when GraphQL serves the read, an SDL pulled from the target org).
Knowledge Bases
Per-adapter API reference — references/adapter-apis.md holds Syntax / Parameters / Returns / Usage for every UI API adapter, grouped by family (uiRecordApis, uiListsApis, uiRelatedListApis, uiObjectInfoApis). Each adapter is a # `<name>` block; grep for the backticked name (e.g. # `getRecord`) to jump to its entry. Read this before wiring an adapter; do not paraphrase from memory.
Type catalog — references/wire-adapter-types.md holds every type the adapters return (Record, ObjectInfo, FieldValue, etc.), grouped by category and rendered with the same formatter the legacy MCP tool used. Grep for ## <TypeName> to jump to a specific entry.
Read the applicable reference before editing code.
Core Principles
- Prefer LDS/UIAPI for CRUD on standard and custom objects. Use Apex only when business logic or bulk operations exceed LDS capabilities.
- Always keep rendered data fresh with
refreshApex(wiredResult) or notifyRecordUpdateAvailable([{ recordId }]) after any mutation.
- Import object and field references from
@salesforce/schema — not string literals. This protects the component against metadata renames.
- Favor base record form components (
lightning-record-form, lightning-record-edit-form, lightning-record-view-form) for single-record UIs. They ship with validation, SLDS styling, accessibility, and field-level security.
Review Checklist
Answer Yes / No to each. Any Yes triggers a refactor.
1. Hand-rolling forms instead of base components
- Does the component implement a custom form for single-record CRUD where
lightning-record-form, lightning-record-edit-form, or lightning-record-view-form would suffice?
- Does validation logic duplicate what base record form components provide natively?
- Are standard SLDS styles recreated manually instead of leveraging the styling baked into base components?
2. Not importing references
- Are object or field API names referenced as hard-coded strings?
- In templates, are field values accessed directly via expressions like
record.data.fields.Name.value without schema imports?
- Does the JS file lack any
@salesforce/schema import even though it interacts with Salesforce fields?
3. Mixing Apex and LDS without synchronization
- Does the component read via LDS and mutate the same record through Apex without a subsequent cache refresh?
- Does it fetch through Apex yet rely on the LDS cache for display without synchronizing after updates?
- Do multiple data sources touch the same object without an explicit refresh strategy?
4. Overusing Apex
- Does the component call Apex solely to retrieve or update a single record that
getRecord, updateRecord, or a base record form could handle?
- Is Apex used to run a simple SOQL query whose fields are available through standard LDS wire adaptors?
- Are custom Apex methods present for basic CRUD while no LDS/UIAPI calls appear in the code?
Workflow
Step 1 — Inventory the data layer
List every data operation in the component:
- Wire adapters (
@wire(getRecord, …), @wire(getRecords, …), @wire(someApexMethod, …)).
- Imperative calls (
updateRecord, createRecord, deleteRecord, Apex imperative).
- Reads vs writes, target object(s), fields, and whether the refresh path after writes is wired.
Step 2 — Run the four-section checklist
Walk sections §1–§4 in order. For each section, decide whether it applies to the component under review and record the result in the report. Use the shape below — every section must appear exactly once, either as an issue (violation) under ## LDS Best Practices, or as a compliant entry under ## Sections checked (no issue). Finish with a ## Summary line listing counts and a one-paragraph narrative.
Report shape:
## LDS Best Practices
- §<N> <section title> — <file>:<lines>
Issue: <what is wrong, specifically citing the pattern in the code>
Fix: <the corrective change, naming the exact import / API to use>
Applied: <yes | no>
## Sections checked (no issue)
- §<N> <section title> — <file>:<lines>
Status: Compliant (no action).
Evidence: <what in the code makes this section — , , >
issue(s) found; fixed; deferred.
Rules for producing this report:
- Every one of §1–§4 must appear in exactly one of the two blocks. Do not omit a section because it is compliant; record it with evidence under
## Sections checked (no issue).
- Under
## LDS Best Practices, only list actual violations. If there are no violations, write "No best-practice issues found." as the first line, then move every section to the compliant block.
- Cite specific file paths and line ranges from the component under review — never generic references.
- Do not invent additional sections beyond §1–§4; downstream a11y / RTL / security reviews run as separate workflows and have their own reports.
Step 3 — Apply referential-integrity fixes
For every hard-coded API name:
import ACCOUNT_OBJECT from '@salesforce/schema/Account';
import NAME_FIELD from '@salesforce/schema/Account.Name';
import INDUSTRY_FIELD from '@salesforce/schema/Account.Industry';
Use these constants everywhere the object or field is referenced — wire configs, @wire field arrays, getFieldValue(record, NAME_FIELD) calls, and base-component object-api-name / fields attributes.
Full rules: references/lds-referential-integrity.md.
Step 4 — Apply data-consistency fixes
- After imperative LDS mutation (
updateRecord, createRecord, deleteRecord), dispatch a refresh:
import { updateRecord, getRecord } from 'lightning/uiRecordApi';
import { refreshApex } from '@salesforce/apex';
async handleSave() {
await updateRecord({ fields: { Id: this.recordId, Name: this.name } });
await refreshApex(this.wiredRecord);
}
- After Apex mutation of a record the cache holds, prefer:
import { notifyRecordUpdateAvailable } from 'lightning/uiRecordApi';
await notifyRecordUpdateAvailable([{ recordId: this.recordId }]);
- Keep a reference to wire results (
this.wiredRecord = result; return result.data;) so refreshApex can target them.
- Base form components refresh themselves; no manual refresh needed.
Full rules: references/lds-data-consistency.md.
Step 5 — Replace Apex with UIAPI where applicable
- Single-record read →
getRecord (with fields + schema imports).
- Single-record update →
updateRecord or lightning-record-edit-form.
- Single-record create →
createRecord or lightning-record-form with mode="edit".
- Related-record read →
getRelatedListRecords.
- Picklist values →
getPicklistValues.
- Object metadata →
getObjectInfo / getObjectInfos.
When in doubt about adapter shape, grep references/adapter-apis.md for the backticked adapter name (e.g. # `getRecord`) — it has the authoritative parameters, returns, and usage. For @salesforce/schema/<Object>.<Field> paths, confirm the exact API name in Setup → Object Manager → <Object> → Details → API Name. For unfamiliar return types, grep references/wire-adapter-types.md for the type name.
Step 6 — Verify
- No hardcoded API names remain in the component files.
- Every write path has a matching refresh path (or uses a base form component).
- No duplicate reads of the same record via both Apex and UIAPI.
- Existing Jest tests pass; add coverage for the refresh flow (
refreshApex called exactly once per mutation).
Cross-References
- Related skills:
experience-lwc-generate — when the review surfaces the need to regenerate rather than patch the component.
design-systems-slds-apply — for SLDS class / design-token cleanup surfaced by the LDS review.
- Adjacent (out-of-band today):
- GraphQL query/mutation authoring, upstream data-requirements analysis, and the security / RTL / a11y review passes run as separate workflows with their own tooling.
Examples
Base-component first (preferred)
<template>
<lightning-record-form
record-id={recordId}
object-api-name="Account"
fields={fields}
mode="edit"
onsuccess={handleSuccess}>
</lightning-record-form>
</template>
import { LightningElement, api } from 'lwc';
import NAME_FIELD from '@salesforce/schema/Account.Name';
import INDUSTRY_FIELD from '@salesforce/schema/Account.Industry';
export default class AccountEditor extends LightningElement {
@api recordId;
fields = [NAME_FIELD, INDUSTRY_FIELD];
handleSuccess() {
this.dispatchEvent(new CustomEvent('saved'));
}
}
Imperative update with refresh
import { LightningElement, api, wire } from 'lwc';
import { getRecord, updateRecord } from 'lightning/uiRecordApi';
import { refreshApex } from '@salesforce/apex';
import ACCOUNT_NAME from '@salesforce/schema/Account.Name';
export default class RenameAccount extends LightningElement {
@api recordId;
wiredRecord;
@wire(getRecord, { recordId: '$recordId', fields: [ACCOUNT_NAME] })
wired(result) {
this.wiredRecord = result;
}
async handleRename(event) {
await updateRecord({ fields: { Id: this.recordId, Name: event.detail } });
await refreshApex(this.wiredRecord);
}
}
Verification
- Grep for
@salesforce/schema/ imports — they should cover every field/object the component references.
- Grep for string literals that look like API names (
'Account', 'Name') — none should appear in wire configs or field arrays.
- Trace every mutation call to a refresh call (either
refreshApex, notifyRecordUpdateAvailable, or a base form handling it internally).
- Confirm Apex is only used where UIAPI can't satisfy the requirement (bulk, complex joins, custom logic).