forcedotcom/sf-skills

experience-lwc-base-components-integrate

Pick the right Lightning Base Component (lightning-) for a given UI task, retrieve its full API (props, methods, events, slots) from the bundled per-component reference, and wire it into an LWC (LWC .html, .js, and .css files) without breaking SLDS.

Ver código-fonte
Documento original do Skill

Renderizado do repositório de origem, preservando títulos, exemplos, código, tabelas, links e imagens.

<!-- adk-managed-skill -->

Using Lightning Base Components

Lightning Base Components (LBC) are the lightning-* web components shipped by Salesforce. This skill routes an agent through the right decision sequence so the final component choice is as specific as possible and backed by real API docs — not a hand-rolled reimplementation of something that already exists.

When to Use This Skill

  • User describes a UI need ("searchable dropdown", "record edit form",

"modal with footer") and asks which lightning-* component fits.

  • User is about to build a primitive (button group, combobox, toast) and

should be using LBC instead.

  • User asks you to review LWC markup for LBC-related issues — specifically

overriding SLDS classes or restyling LBC internals.

  • User needs the authoritative props/events/slots for a specific

lightning-* tag.

Prerequisites

  • Knowledge of which LBC namespace your org uses (lightning is the default

public namespace; some platforms expose lightning-community or others — the user's meta files will clarify).

  • The skill ships authoritative API docs for every Lightning Base Component

in references/lightning-components.md. Each component is a # Component API Structure block; grep for **Name:** <camelCaseName> (e.g. **Name:** datatable) to jump to its Properties / Methods / Events / Slots. Read this rather than relying on cached knowledge — LBC evolves and the reference is the source of truth.

Workflow

Step 1 — Read the entire component index first

Open lightning-component-index.md and scan all entries before making any selection. This is non-negotiable: LBC's value comes from picking the most specialized component, and skipping the scan leads to reinventing compound widgets out of primitives.

As you scan, compile a candidate list — every component whose description touches any aspect of the use case. Do not filter or rank yet.

Step 2 — Narrow to the most specific fit per feature

Once the scan is complete:

  • For each feature in the use case, select the most specific component

that covers it. Prefer a specialized compound (lightning-record-form, lightning-tabset, lightning-datatable) over a generic primitive (lightning-input, lightning-button) when the specialized one covers the scenario end-to-end.

  • Avoid duplication: if lightning-record-form already renders fields for a

record, do not pair it with lightning-input-field unless you're explicitly overriding behavior.

Step 3 — Share the shortlist and confirm

Present the final shortlist to the developer with a one-line rationale per component. Wait for explicit confirmation before pulling full API docs. This prevents the agent from burning context on components the developer has already mentally ruled out.

Step 4 — Retrieve full API docs

Once confirmed, use the bundled helper to pull the exact API blocks — this avoids ad-hoc grepping across a large reference:

bash
scripts/extract-component-docs.sh <camelCaseName> [<camelCaseName>...]

Convert lightning-<foo> tags to camelCase (no lightning- prefix):

  • lightning-datatable → datatable
  • lightning-record-edit-form → recordEditForm
  • lightning-button-icon → buttonIcon

Each returned block has the same shape: Basic Information (tag, namespace, type), Properties (name, type, default, description), Methods, Events, Slots, and (where applicable) usage notes. This skill is about picking the components; the bundled reference is about wiring them.

Step 5 — Produce integration guidance

Using the per-component reference, walk the developer through:

  • The exact <lightning-...> tag and required attributes.
  • Which events to bind (onchange, oncommit, onsuccess, …) and what

the event payload contains.

  • Any slots to fill (headers, footers, custom content).
  • Known constraints from the component docs (e.g. lightning-record-form

requires object-api-name and record-id for edit/view modes).

Step 6 — Respect LBC styling rules

Do not override SLDS classes on LBC internals. See lbc-expert-guidance.md for specifics. Common issues:

  • Targeting .slds-button or .slds-input in the host component's CSS to

restyle an LBC — LBC ships inside a shadow root, so these selectors either leak into sibling components or get stripped entirely. Use the component's documented styling hooks (--sds-c-button-*, etc.) instead.

  • Wrapping an LBC just to mutate its internal markup. You can't — the

markup is hidden behind the shadow root. If the component doesn't expose the slot/prop you need, that's a platform-level gap, not a restyling job.

Examples

Example — "I need a multi-select combobox with typeahead"

  1. Scan the component index end-to-end.
  2. Candidate list includes: lightning-combobox, lightning-dual-listbox,

lightning-record-picker.

  1. Shortlist: lightning-dual-listbox (the documented multi-select base

component). Rule out lightning-combobox — its documented API is single-select; it has no type="multi" and no multi-select mode. Flag that lightning-record-picker only fits if the values are record IDs.

  1. Developer confirms lightning-dual-listbox.
  2. Run scripts/extract-component-docs.sh dualListbox.
  3. Return the options, value, onchange payload, and required label

props from the block's Properties / Events sections.

Example — "I'm going to write my own modal"

  1. Scan finds lightning-modal, lightning-modal-body,

lightning-modal-footer, lightning-modal-header.

  1. Shortlist is the 4 modal components.
  2. Confirm.
  3. Run scripts/extract-component-docs.sh modal modalHeader modalBody modalFooter

→ full modal API (how to extend LightningModal, the static .open() pattern, slotting the header/body/footer).

  1. Steer the developer away from rolling their own dialog.

Verification Checklist

  • [ ] The full component index was scanned before any selection (no

keyword-search shortcutting).

  • [ ] Candidate list included every component that touches the use case.
  • [ ] Final shortlist selects the most specific component per feature.
  • [ ] Developer confirmed the shortlist before the bundled component

reference was opened.

  • [ ] Integration guidance cites props / events / slots from the real API

docs (not inferred).

  • [ ] No suggestions to restyle LBC by overriding SLDS classes.

Troubleshooting

  • Grep for `Name: <name>` returns no match — name is wrong, or the

reference uses a different camelCase. Double-check against the component index (lightning-record-form → recordForm, lightning-record-view-form → recordViewForm, lightning-button-icon → buttonIcon).

  • Proposed component doesn't have the prop you expected — trust the

real API doc over memory. LBC evolves; cached knowledge lies.

  • Developer resists the shortlist — don't skip Step 4. Still retrieve

the docs for the developer's preferred choice so they see the actual trade-offs.

  • Developer wants to restyle LBC internals — redirect to styling hooks

(see the LBC Expert reference). Refusing shadow DOM penetration is the correct answer.

do mesmo repositório

Mais Skills

Todos os Skills
forcedotcom
Comunidade

agentforce-generate

Build, modify, audit, repair, optimize, debug, and deploy agents with Agentforce Agent Script. TRIGGER when: user creates, reviews, or changes .agent files or aiAuthoringBundle metadata; asks to fix AgentScript, audit an existing agent, run an AgentScript health check, common-pitfall review, or baseline-versus-candidate repair loop; changes a response, action, subagent, route, state flow, or Agent Spec; migrates or upgrades a legacy GenAiPlannerBundle Agentforce agent (NOT an Einstein Bot) to an

instalações
1
GitHub Stars
1,1 mil
Atualizado
7 de out.
forcedotcom
Comunidade

agentforce-test

Write, run, and analyze structured test suites for Agentforce agents — functional AND security. TRIGGER when: user writes or modifies test spec YAML (AiEvaluationDefinition); runs sf agent test create, run, run-eval, or results commands; asks about test coverage strategy, metric selection, or custom evaluations; interprets test results or diagnoses test failures; asks about batch testing, regression suites, or CI/CD test integration; requests security testing, OWASP LLM Top 10, red-teaming, pene

instalações
1
GitHub Stars
1,1 mil
Atualizado
7 de out.
forcedotcom
Comunidade

automation-flow-generate

Generate Salesforce Flows using the MCP tool executemetadataaction. Use when the user asks to create, build, or generate a flow — including Screen, Autolaunched, Record-Triggered (before/after-save), Scheduled. Also trigger for flow-like requests such as \"when a record is created\", \"trigger daily at\", \"send an email when\", \"update the field when\", \"automate\", \"workflow\", or \"flow XML/metadata\". This is the only skill for Salesforce Flow generation.

instalações
1
GitHub Stars
1,1 mil
Atualizado
7 de out.
forcedotcom
Comunidade

dx-code-analyzer-configure

Set up, configure, and troubleshoot Salesforce Code Analyzer for any project. Handles installation, prerequisite checks, diagnosing broken setups, creating and editing code-analyzer.yml overrides, engine-specific settings, ignore patterns, severity overrides, and CI/CD pipeline setup. TRIGGER when: user says 'set up code analyzer', 'configure code analyzer', 'install code analyzer', 'code analyzer not working', 'fix my setup', 'scan failing', 'check my setup', 'enable/disable engine', 'exclude f

instalações
1
GitHub Stars
1,1 mil
Atualizado
7 de out.