AI-powered Salesforce development tools for Claude Code. 23 agents, 9 commands, 6 skills — all designed to work together in a sequential workflow.
A plugin that gives Claude deep Salesforce expertise for:
- Code reviews that catch governor limits, security issues, and anti-patterns
- Implementation that follows best practices from the start
- Testing with proper bulk scenarios and assertions
- Deployment with checklists and validation
Core Philosophy: Each unit of work should make subsequent work easier—not harder.
# Clone the repository
git clone https://github.com/gellasangameshgupta/sf-compound-engineering-marketplace.git
# Install in Claude Code
/plugin install /path/to/sf-compound-engineering-marketplaceThe real power is using commands in sequence. Here's a complete example:
Requirement:
Build a Lead scoring system that calculates scores based on lead source, industry, and company size. Update scores on lead changes. Sync high-scoring leads (>80) to external marketing system via REST API. Send email alerts for hot leads.
/sf:plan
I need to build a Lead scoring system that:
- Calculates a score based on lead source, industry, and company size
- Updates the score whenever a lead is modified
- Syncs high-scoring leads (>80) to an external marketing system via REST API
- Sends email alerts to sales reps for hot leads
What Claude Does:
- Breaks down into phases
- Identifies components needed (trigger, handler, service classes)
- Estimates complexity
- Lists dependencies
Output: Implementation plan you'll reference in next steps
/sf:work
Implement Phase 1 from the plan: Lead Score Calculation
Create:
1. Lead_Score__c field (Number)
2. LeadTrigger
3. LeadTriggerHandler
4. LeadScoringService with calculateScore() method
What Claude Does:
- Writes bulk-safe Apex code
- Follows trigger handler pattern
- Includes CRUD/FLS checks
- References
apex-patternsskill automatically
Repeat for each phase:
/sf:work
Implement Phase 2: External System Sync (MarketingSystemService, AsyncLeadSyncJob)
/sf:work
Implement Phase 3: Email Alerts (LeadAlertService)
/sf:review
Review all the Lead Scoring code:
- LeadTrigger.trigger
- LeadTriggerHandler.cls
- LeadScoringService.cls
- MarketingSystemService.cls
What Claude Does:
- Dispatches multiple agents automatically:
apex-governor-guardian→ checks limitsapex-security-sentinel→ checks CRUD/FLS, injectionapex-bulkification-reviewer→ checks bulk patternscallout-pattern-reviewer→ checks HTTP patterns
- Consolidates findings by severity
Output:
### [CRITICAL] Missing CRUD Check
Location: LeadScoringService.cls:45
...
### [HIGH] Callout Missing Timeout
Location: MarketingSystemService.cls:23
...
/sf:triage
Process the review findings. Help me decide what to fix now vs later.
What Claude Does:
- Presents each finding interactively
- Helps categorize: Fix Now / Defer / Skip
- Creates prioritized todo list
Output:
Accepted (Fix Now):
- [ ] CRITICAL: Add CRUD check in LeadScoringService
- [ ] HIGH: Add timeout to MarketingSystemService
Deferred (Backlog):
- [ ] MEDIUM: Add retry logic to callout
Skipped:
- LOW: Variable naming suggestion (existing pattern)
/sf:resolve critical,high
Fix all critical and high priority issues from triage.
What Claude Does:
- Applies fixes to each issue
- Shows before/after code
- Validates fixes don't break other things
/sf:test
Generate tests for the Lead Scoring feature:
- LeadScoringService (unit tests)
- LeadTriggerHandler (bulk tests with 200+ records)
- MarketingSystemService (mock callout tests)
Target: 90% coverage
What Claude Does:
- Creates test classes using
test-factoryskill - Includes bulk scenarios (200+ records)
- Adds mock classes for callouts
- Covers positive, negative, and edge cases
/sf:document
Document the Lead Scoring feature:
- ApexDoc for all classes
- Feature README
What Claude Does:
- Adds ApexDoc comments to all methods
- Creates feature documentation
- Documents architecture decisions
/sf:health
Analyze Lead Scoring codebase health before deployment.
What Claude Does:
- Reports code coverage
- Identifies remaining technical debt
- Scores pattern compliance
- Gives go/no-go recommendation
/sf:deploy production
Deploy Lead Scoring feature:
- All Apex classes and triggers
- Lead_Score__c custom field
- Named Credential for Marketing API
What Claude Does:
- Generates pre-deployment checklist
- Provides SF CLI deployment commands
- Creates post-deployment verification steps
- Documents rollback procedure
┌─────────────────────────────────────────────────────────────┐
│ │
│ /sf:plan │
│ │ │
│ ▼ │
│ /sf:work ◄─────────────────┐ │
│ │ │ │
│ ▼ │ │
│ /sf:review │ │
│ │ │ │
│ ▼ │ │
│ /sf:triage │ Iterate until │
│ │ │ review is clean │
│ ▼ │ │
│ /sf:resolve ───────────────┘ │
│ │ │
│ ▼ │
│ /sf:test │
│ │ │
│ ▼ │
│ /sf:document │
│ │ │
│ ▼ │
│ /sf:health │
│ │ │
│ ▼ │
│ /sf:deploy ──────────────────► PRODUCTION 🚀 │
│ │
└─────────────────────────────────────────────────────────────┘
Specialized AI personas that know Salesforce inside out:
| Category | Agents | What They Check |
|---|---|---|
| Apex (6) | governor-guardian, security-sentinel, bulkification-reviewer, trigger-architect, test-coverage-analyst, exception-handler | Limits, CRUD/FLS, injection, bulk patterns, test quality |
| LWC (5) | performance-oracle, security-reviewer, accessibility-guardian, architecture-strategist, aura-migration-advisor | Wire adapters, XSS, ARIA, component patterns |
| Automation (4) | flow-complexity-analyzer, flow-governor-monitor, process-automation-strategist, validation-rule-reviewer | Flow design, DML in loops, Flow vs Apex decisions |
| Integration (4) | rest-api-architect, callout-pattern-reviewer, platform-event-strategist, integration-security-sentinel | API design, Named Credentials, retry patterns |
| Architecture (4) | data-model-architect, sharing-security-analyst, metadata-consistency-checker, pattern-recognition-specialist | Schema design, OWD, anti-patterns |
| Command | Purpose | When To Use |
|---|---|---|
/sf:plan |
Create implementation plan from requirements | Start of any new feature |
/sf:work |
Execute plan with best practices | Building components |
/sf:review |
Multi-agent code review | After writing code |
/sf:triage |
Prioritize review findings | After review |
/sf:resolve |
Fix prioritized issues | After triage |
/sf:test |
Generate comprehensive tests | Before deployment |
/sf:document |
Auto-generate documentation | After code is stable |
/sf:health |
Analyze codebase health | Before major deploys |
/sf:deploy |
Deployment checklists & commands | Ready to ship |
Reference documentation Claude consults automatically:
| Skill | What It Contains |
|---|---|
governor-limits |
All Salesforce limits with thresholds |
security-guide |
CRUD/FLS patterns, injection prevention |
apex-patterns |
Trigger handler, selector, service patterns |
lwc-patterns |
Component communication, state management |
test-factory |
TestDataFactory patterns |
integration-patterns |
REST, callouts, platform events |
/sf:review
public class MyService {
public void process(List<Account> accs) {
for (Account a : accs) {
Contact c = [SELECT Id FROM Contact WHERE AccountId = :a.Id LIMIT 1];
update a;
}
}
}
Expected: Claude flags SOQL in loop, DML in loop, missing CRUD check
/sf:plan
Add a "Convert to Customer" button on Account that:
- Creates an Opportunity
- Sends welcome email
- Updates Account type to "Customer"
/sf:test
Generate bulk tests for AccountTriggerHandler class.
Target 90% coverage.
The plugin provides Claude with:
-
Auto-Dispatch Rules (in CLAUDE.md)
- When Claude sees
.clsfiles → automatically uses Apex agents - When Claude sees
.jsin lwc/ → automatically uses LWC agents - When asked about limits → automatically reads governor-limits skill
- When Claude sees
-
Structured Prompts (agents)
- Each agent has expertise, checklists, response format
- Claude adopts the agent's persona for specialized reviews
-
Reference Knowledge (skills)
- Curated Salesforce best practices
- Copy-paste ready code patterns
- Limit thresholds and security guides
sf-compound-engineering-marketplace/
├── .claude-plugin/
│ └── marketplace.json # Plugin registry
├── plugins/
│ └── sf-compound-engineering/
│ ├── .claude-plugin/
│ │ └── plugin.json # Plugin metadata
│ ├── agents/
│ │ ├── apex/ # 6 Apex agents
│ │ ├── lwc/ # 5 LWC agents
│ │ ├── automation/ # 4 Flow agents
│ │ ├── integration/ # 4 Integration agents
│ │ └── architecture/ # 4 Architecture agents
│ ├── commands/ # 9 slash commands
│ ├── skills/ # 6 reference skills
│ └── CLAUDE.md # Auto-dispatch rules
├── README.md
├── CHANGELOG.md
├── CONTRIBUTING.md
└── LICENSE
cd /path/to/sf-compound-engineering-marketplace
git pull origin main
# If changes don't take effect:
/plugin uninstall sf-compound-engineering
/plugin install /path/to/sf-compound-engineering-marketplaceSee CONTRIBUTING.md for guidelines on adding new agents, commands, or skills.
MIT License - see LICENSE
- Inspired by Every.to's Compound Engineering
- Built for the Salesforce developer community
- Powered by Claude Code
Made with ❤️ for Salesforce Developers