Saltar al contenido
Marketplace

experience-aura-lwc-migrate

de forcedotcom

Use this skill to analyze a Salesforce Aura component bundle (.cmp, .app, .evt, .intf, Controller.js, Helper.js, Renderer.js) and produce a framework-agnostic migration blueprint (PRD.yaml / PRD.md / PRD.json) capturing public API, data requirements, slots, events, states, accessibility, styling, localization, and security posture. This is Phase 1 (analysis + PRD) only — it does NOT author or edit component code, wire adapters, or Jest tests, and does not write .html/.js/.css/.js-meta.xml. TRIGGER when the user says \"analyze this Aura component\", \"migrate Aura\", \"convert .cmp\", \"produce a migration blueprint\", \"generate a PRD for this Aura component\", or mentions aura:attribute, aura:handler, force:recordData, $Label, $Resource, component.get, cmp.find, or an input directory containing .cmp files. DO NOT TRIGGER when there is no Aura source (delegate downstream LWC authoring to experience-lwc-generate), for post-migration scoring, or for refactoring an existing modern component.

Resumen

SKILL.mdLeída en abb06a4

Migrating Aura to LWC

Migrate an existing Salesforce Aura component bundle to a Lightning Web Component using a PRD-driven workflow. The Aura source is analyzed through eight expert lenses (static references, data, API surface, slots, styling, default values, unknown dependencies, redundant-code cleanup) to produce a framework-agnostic PRD, which is handed to the LWC authoring skills and verified against the Aura→LWC completeness checklist.

When to Use

  • Porting an Aura bundle (.cmp, .controller.js, .helper.js, .renderer.js, .css, .design, .evt, .intf) to LWC.
  • Producing a framework-agnostic PRD from Aura source so a sibling team can implement the LWC equivalent.
  • Auditing an Aura component for patterns that need explicit mapping before LWC conversion (facets, aura:id DOM access, force:recordData, application events).

Do NOT use this skill for:

  • Authoring a brand-new LWC from a Figma or PRD with no Aura source (delegate to experience-lwc-generate).
  • Refactoring an existing LWC (delegate to experience-lwc-generate).
  • Post-hoc completeness scoring of a conversion that was already done without this skill, or Lightning Out Beta → 2.0 host-page migration — no checked-in skill covers these today.

Prerequisites

  • Aura component bundle files (minimum: .cmp; usually also controller / helper / CSS / events / interfaces).
  • Target LWC path (module folder) — discover the caller's module tree from sfdx-project.json packageDirectories (typically an lwc/<name>/ subdirectory under one of the listed package paths).
  • For resolver-assisted dependency lookup (Phase 2, step 7): the @sfdc-internal/adk-knowledge package installed in the caller's workspace. If it is not present, the skill surfaces the missing dependency as an unknown rather than failing.
  • Awareness of the project's downstream LWC authoring flow — this skill hands off a PRD to that flow but does not invoke it.

Core Principles

Apply the Aura → LWC migration principles before and during each phase: understand before converting, prioritize functionality equivalence over structural equivalence, leverage native web standards, migrate iteratively, and test thoroughly. The architectural deltas (two-way → one-way binding, Aura events → DOM events, aura:id → this.template.querySelector, etc.) are documented in references/aura-migration-guidelines.md.

Knowledge Bases

Open the relevant reference before drafting or enhancing the corresponding PRD section.

Workflow

This skill covers the analysis and PRD-generation phases only. It does not produce, edit, or validate LWC bundles, wire adapters, or tests. Phases 1–3 below are mandatory when this skill is invoked. Phases 4–5 describe the recommended downstream authoring flow — they are informational follow-up guidance for the caller and are NOT executed by this skill.

Phase 1 — Analyze and draft the PRD

Goal: produce a draft YAML PRD that captures every section of the 14-section framework.

  1. Inspect the caller-supplied Aura bundle files. Extract the component name, the access attribute from the .cmp metadata, and every embedded/referenced component.
  2. Apply the framework in references/aura-prd-framework.md. State the access attribute explicitly in the overview and describe its implications (GLOBAL: publicly available across namespaces; PRIVILEGED/PUBLIC: restricted; etc.).
  3. Never list the host component itself in the unknowns section — that creates a circular dependency. If no valid unknowns exist, use an empty array.
  4. Save the draft PRD to the workspace (e.g., packages/skills/<skill>-workspace/<iteration>/PRD-draft.yaml).

Deliverable: draft PRD (YAML) covering the 14 sections.

Phase 2 — Enhance the PRD with Aura expert lenses

Goal: harden each section using the eight Aura expert frameworks.

Apply the experts in order. Each pass reads the draft PRD and rewrites the relevant sections in place.

  1. Reference analysis — apply references/aura-reference-expert.md to staticReferences, componentCommunication, and unknowns. Enumerate every $Resource, $ContentAsset, $Label, $Browser, $Locale, /lightning/* route, and c:* / force:* / lightning:* dependency. The blueprint must never directly contain the Aura-specific strings $ContentAsset, $Label, $Resource, $Browser, or $Locale; replace each with a description of intent.
  2. Data analysis — apply references/aura-data-expert.md to dataRequirements. The blueprint dataRequirements must not contain force:recordData or lightning:recordViewForm — document the underlying record, fields, and object-API-name instead. Rename sObjectName → objectApiName. Capture Apex-controller method names, input parameters, and output shapes.
  3. API analysis — apply references/aura-api-expert.md to componentCommunication, interactions, and staticReferences. Document every public attribute (with type, default, and all usage intents), <aura:method> declaration, event firing (component.getEvent(...).fire()), and Object-typed callback attribute (these become LWC custom events).
  4. Slots analysis — apply references/aura-slots-expert.md to contentRequirements. Translate every {!v.body} / {#v.body} to a default slot entry (LWC components can have only one default slot — capture the condition under which it is present) and every <aura:attribute type="Aura.Component"> or "Aura.Component[]" to a named-slot entry.
  5. Style analysis — apply references/aura-style-expert.md to styling and dataRequirements. Move styling-related attributes out of dataRequirements and into styling. Document SLDS hooks, custom CSS, and any dynamic styling driven by JS.
  6. Values analysis — apply references/aura-values-expert.md to dataRequirements. For every Aura attribute, record its default value (type + literal; or namespace + name for $Label references) in the what property.
  7. Resolver analysis — apply references/aura-resolver-expert.md to unknowns. For each unknown, parse namespace + name, look up the corresponding file under node_modules/@sfdc-internal/adk-knowledge/dist/ using the mapping rules, and either inline the resolved content or leave a note that manual research is required. If the @sfdc-internal/adk-knowledge package is not installed, note the expected path and leave the unknown in place.
  8. Redundant-code cleanup — apply references/aura-redundant-code-expert.md across the PRD. Remove commented-out markup/JS/CSS, telemetry/instrumentation, and private attributes/methods that aren't used elsewhere — these don't carry forward to LWC. Keep the PRD's JSON shape intact; sections may end up with empty arrays. Generate a short report listing only the items that actually changed.

Deliverable: enhanced PRD (YAML) with all eight expert passes applied.

Phase 3 — Accessibility pass

Goal: ensure the PRD's accessibility section meets WCAG 2.2 minimums.

  1. Apply the WCAG 2.2 reviewers inline against the Aura source files (.cmp, .controller.js, .helper.js, .renderer.js, .css). Cover the Aura-scoped Success Criteria: 1.3.1 (ii) Tables, 1.3.5 Identify Input Purpose, 3.2.1 On Focus, 3.2.2 On Input, 3.3.2 Labels or Instructions, 4.1.2 Name / Role / Value. (When the sibling accessibility-code-review-aura skill ships in a follow-up PR, hand off to it instead.)
  2. Fold each finding into the PRD's accessibility section. For every requirement, capture the WCAG Success Criterion it satisfies and the proposed LWC implementation approach (semantic HTML first, ARIA only where semantic HTML cannot achieve the same outcome).
  3. Only cite violations that directly break a WCAG Success Criterion. Do not suggest enhancements outside WCAG scope.

Deliverable: PRD with an accessibility section that lists each required feature, the WCAG Success Criterion it satisfies, and the LWC implementation approach.

Phase 4 — Downstream authoring (follow-up guidance, not executed here)

Once the enhanced PRD is complete, the caller can hand it off to an LWC authoring workflow. This skill does not perform any authoring. Recommended downstream steps:

  • Pass the enhanced PRD to experience-lwc-generate as the design input, substituting for a Figma / written-spec input. This skill does not invoke experience-lwc-generate; the caller triggers it as a follow-up step.
  • For data access decisions (UIAPI / GraphQL / Apex), the caller may consult LDS-focused workflows to translate the PRD's dataRequirements section into concrete adapters.
  • For SLDS styling decisions, the caller may consult an SLDS design workflow.
Phase 5 — Downstream verification (follow-up guidance, not executed here)

After LWC authoring completes, the caller can verify the migrated component with the project's standard test and review workflows — Jest coverage, accessibility Jest, a11y / security / RTL review, telemetry, feature-flag gating, etc. Apply references/aura-to-lwc-completeness-checklist.md as a final scoring pass on the completed LWC.

Cross-References

  • experience-lwc-generate — recommended downstream authoring handoff (non-executing; the caller triggers it after this skill's PRD is complete).

Verification (PRD-only)

  • The enhanced PRD covers all 14 framework sections with no TODO / placeholder entries.
  • dataRequirements contains no direct references to force:recordData or lightning:recordViewForm, and no sObjectName keys (renamed to objectApiName).
  • staticReferences and other sections contain no raw Aura strings like $Resource, $Label, $ContentAsset, $Browser, $Locale.
  • Every <aura:method> appears in interactions with the corresponding trigger.
  • Every {!v.body} / {#v.body} appears in contentRequirements as a default slot (with the presence condition).
  • Every <aura:attribute type="Aura.Component"> appears in contentRequirements as a named slot.
  • Every unknown has been resolved via the @sfdc-internal/adk-knowledge package (or marked as manually-required with the attempted knowledge-file path).

Archivos 12 archivos

Instálala a mano

Deska ejecuta esto por ti, para cada agente que elijas. Para ejecutarlo tú en una terminal:

npx skills add forcedotcom/sf-skills --skill experience-aura-lwc-migrate

Más de forcedotcom

  • platform-apex-generateforcedotcomPrimary Apex authoring skill for class generation, refactoring, and review. ALWAYS ACTIVATE when the user mentions Apex, .cls, triggers, or asks to create/refactor a class (service, selector, domain, batch, queueable, schedulable, invocable, DTO, utility, interface, abstract, exception, REST resource). Use this skill for requests involving SObject CRUD, mapping collections, fetching related records, scheduled jobs, batch jobs, trigger design, @AuraEnabled controllers, @RestResource endpoints, custom REST APIs, or code review of existing Apex.6,700 instalaciones en skills.sh0
  • platform-apex-test-generateforcedotcomUse to generate and validate Apex test classes with TestDataFactory patterns, bulk testing (251+ records), mocking, assertions, and disciplined test-fix loops. Use when creating Apex test classes (for triggers, services, controllers, batch jobs, queueables, and callouts), improving coverage, debugging or fixing failing Apex tests, or running test/coverage analysis. Triggers on *Test.cls files, sf apex run test, coverage reports. Do NOT trigger for production Apex (use platform-apex-generate) or Jest/LWC tests.6,600 instalaciones en skills.sh0
  • platform-apex-test-runforcedotcomApex test execution, coverage analysis, and test-fix loops with 120-point scoring. Use when the user runs Apex tests, checks code coverage, fixes failing tests, or touches *Test.cls / *_Test.cls files. DO NOT TRIGGER when writing Apex production code (use platform-apex-generate), Agentforce agent testing (use agentforce-test), or Jest/LWC tests (use experience-lwc-generate).6,500 instalaciones en skills.sh0
  • platform-apex-logs-debugforcedotcomSalesforce debug log analysis and troubleshooting with 100-point scoring. TRIGGER when: user analyzes debug logs, hits governor limits, reads stack traces, or touches .log files from Salesforce orgs. DO NOT TRIGGER when: running Apex tests (use platform-apex-test-run), generating or fixing Apex code (use platform-apex-generate), or Agentforce session tracing (use agentforce-observe).6,400 instalaciones en skills.sh0
  • platform-metadata-deployforcedotcomSalesforce DevOps automation using sf CLI v2. TRIGGER when: user deploys metadata, creates/manages scratch orgs or sandboxes, sets up CI/CD pipelines, or troubleshoots deployment errors with sf project deploy. DO NOT TRIGGER when: writing Apex code (use platform-apex-generate), building LWC components (use experience-lwc-generate), creating metadata definitions (use platform-custom-object-generate or platform-custom-field-generate), or querying org data (use platform-data-manage).6,300 instalaciones en skills.sh0
Todas las skills de forcedotcom