Experience Lds Data Requirements Generate

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

Generating LDS Data Requirements

Run a three-stage analyst workflow — requirements clarification, API name validation, API recommendation — so a downstream developer can implement a Lightning Data Service (LDS) solution without guessing.

When to Use

• A PRD, Figma comment, or user ask mentions Salesforce data but objects/fields/operations are vague ("show customer info", "update the record", "list upcoming gigs").
• Before writing any @wire/Apex code for a new data need, or before handing the recommendation to a downstream implementation workflow.
• You inherited TODOs like // TODO: fetch related records and need to turn them into precise specs.

Do NOT use this skill when:

• The data need is already fully specified (object API name, field API names, operation type, scope).
• The component does not touch Salesforce data at all (UI-only, external REST, local state).

Prerequisites

• The natural-language requirement (PRD snippet, user ask, or TODO comment).
• Access to the target org's Setup → Object Manager for confirming custom object/field API names.
• Awareness of the current GraphQL / UI API / Apex priority order (top-of-funnel is GraphQL when it can serve the read).

Knowledge Bases

• references/requirements-analysis.md — Requirements Analysis Mode framework.
• references/api-name-validation.md — Precision Mode for object/field API names.
• references/api-recommendation.md — GraphQL → UI API → Apex decision framework.
• references/lds-expert.md — overall LDS patterns and pitfalls.
• references/lds-data-consistency.md — cache and consistency guarantees.
• references/lds-referential-integrity.md — parent/child and related-record rules.

Workflow

Run the three steps strictly in order. Do not skip a step unless the caller has already confirmed its output.

Step 1 — Parse data requirement (Requirements Analysis Mode)

Goal: extract everything you know and surface every uncertainty before moving on.

Open every conversation with:

"I've analyzed your data requirement: '<REQ>'. Here's what I understand and what I need clarification on…"

Apply the four actions from references/requirements-analysis.md:

1. Operation type — Is it read, create, update, or delete? Ambiguous verbs trigger an immediate clarifying question. Confirm with: "I've identified this as a <OP> operation. Is this correct?"
2. Data entity identification — Standard object (high confidence, proceed), suspected custom object (ask: "Is this a custom object <Term>__c? What's the exact API name?"), or unknown (ask for the API name from Object Manager).
3. Field specification — Map generic references (phone, address, name, status) to specific API names. If multiple candidates exist, enumerate them and ask.
4. Scope and context — One record vs. many; user-triggered vs. auto; expected volume; real-time vs. on-demand.

End-of-step gate. Consolidate into:

Clear Requirements: [confirmed facts]
Need Clarification: [numbered questions from 1.1–1.4]

Proceed only when every question is answered with ≥90% confidence.

Step 2 — Validate Salesforce API names (Precision Mode)

Goal: 100% accuracy on every object and field API name before code is written.

Apply the validation framework from references/api-name-validation.md:

• Standard objects — Account, Contact, Lead, Opportunity, Case, User, Task, Event, Product2, Pricebook2, Order, OrderItem, Asset, Contract, Campaign pass immediately. Anything else triggers verification.
• Custom objects — Never assume __c suffixes. Ask: "Is this <Term>__c or a different custom object API name?" Point users to Setup → Object Manager → <Object> → Details → API Name.
• Standard fields — Map ambiguous references using the tables in the reference file.
• Custom fields — Require __c suffix confirmation; case-sensitive.

Confirmation template:

Object API Name: <OBJECT>
Field API Names: <FIELD_LIST>
Confidence Level: 100% validated

If any uncertainty remains, stop. Emit the outstanding verification requests and the Setup navigation instructions. Do not advance to Step 3 or generate code.

Skip this step only when the caller has explicitly stated that API names are already validated upstream, or the requirement does not involve records at all. Record the skip reason in the Step 4 output.