UNPKG

major-ai-skills

Version:

Installable agentic skills / AI agent skills (SKILL.md) for Claude Code, Cursor, Codex CLI, Gemini CLI & Antigravity - 402+ professional app, token-efficiency, and common-sense skills. SEO/GEO ready.

91 lines (69 loc) 3.29 kB
--- name: salesforce description: "Build Salesforce workflows using SOQL, Apex, Flows, Lightning Web Components, and scoped API integrations." category: enterprise risk: safe source: self source_type: self date_added: "2026-08-26" tags: ["salesforce", "soql", "apex", "flows", "crm", "claude"] tools: ["claude", "cursor", "gemini", "codex"] --- # Salesforce CRM Platform AI Skill Guide ## Overview & Engine Architecture Salesforce is a multi-tenant CRM platform built around standard and custom objects, declarative automation (Flows), and programmatic layers (Apex, LWC). Data is queried with SOQL/SOSL; external systems use REST/Bulk APIs with OAuth. Agents prefer declarative solutions first, keep governor limits in mind, and never hard-code org IDs or secrets in Apex. ``` UI (Lightning / LWC) -> Objects + Fields + Relationships -> Flows / Process automation -> Apex / Triggers (when needed) External apps -> OAuth + REST / Bulk / Streaming APIs ``` ## When to use this skill - Modeling custom objects and relationships for CRM processes - Writing safe SOQL and bulk-friendly Apex - Designing Flows vs Apex trade-offs - Integrating Salesforce with middleware or custom backends ## Operational directives 1. Prefer Flow for straightforward create/update/routing; use Apex for complex bulk logic or callouts. 2. Bulkify everything: no SOQL/DML inside loops. 3. Enforce FLS and sharing (`with sharing` where appropriate); never assume admin context in integrations. 4. Use Named Credentials for callouts; store secrets outside source. 5. Version metadata with SFDX/source format; avoid editing production directly. ## SOQL sketches ```sql SELECT Id, Name, Account.Name FROM Contact WHERE Email = :email LIMIT 1 SELECT Id, (SELECT Id, Subject FROM Cases__r) FROM Account WHERE Id = :accountId ``` ## Apex trigger pattern ```apex trigger ContactTrigger on Contact (before insert, before update) { ContactService.beforeSave(Trigger.new, Trigger.oldMap); } // Service class: one SOQL, one DML path, unit-tested with Test.startTest/stopTest ``` ## Common pitfalls | Pitfall | Result | Fix | | --- | --- | --- | | SOQL in for-loop | Governor limit exceptions | Collect IDs, query once, map results | | Hard-coded record type IDs | Breaks across sandboxes | Query by DeveloperName | | Unrestricted guest/API user | Data leaks | Least-privilege profiles + permission sets | | Flow recursion | Infinite updates | Entry criteria + recursion guards | ## Best practices - Name custom APIs with `__c` awareness; document External IDs for upserts. - Use Bulk API 2.0 for large data jobs; respect daily limits. - Cover Apex with tests that assert behavior, not just coverage percentage. - Prefer Change Data Capture or Platform Events over polling for near-real-time sync. ## Limitations - Edition, licenses, and feature gates (e.g. Flow types) vary by org. - Managed packages and AppExchange apps add opaque automation. - Multi-org middleware patterns need explicit identity and conflict rules. ## Related skills - `@hubspot` - lighter CRM alternative for marketing/sales teams - `@zapier` - no-code sync into Salesforce - `@postman` - explore Salesforce REST collections