Jaganpro/sf-skills provides a comprehensive, community-driven library of specialized skills and agents designed to empower AI coding tools with deep Salesforce development knowledge. It addresses the need for AI agents to understand and generate accurate code for Apex, Flow, LWC, SOQL, Data Cloud, Agentforce workflows, and DevOps practices. This project benefits Salesforce developers and AI researchers by significantly enhancing the capabilities of tools like Claude Code, enabling more efficient and reliable AI-assisted Salesforce development.

How It Works

The core of the project is a collection of over 33 reusable "skills," categorized by functional areas such as development, quality, integration, and AI automation. These skills are integrated with a shared hook system that provides automated validation, linting (using tools like PMD, ESLint), code formatting (Prettier), and Language Server Protocol (LSP) integration for real-time syntax checking and auto-fixing. Seven specialized AI agents are also included, acting as FDE, PS, QA, and Release engineers to orchestrate complex tasks. This layered approach ensures that AI interactions with Salesforce code are context-aware, validated, and adhere to best practices.

Quick Start & Requirements

• Primary Install: For the full experience (skills, agents, hooks, LSP), use the bash installer: curl -sSL https://raw.githubusercontent.com/Jaganpro/sf-skills/main/tools/install.sh | bash. Alternatively, npx skills add Jaganpro/sf-skills installs skills only. A Python installer is available for manual/CI setups.
• Prerequisites: Node.js 18+, Python 3.10+, Salesforce CLI (sf), and an authenticated Salesforce org are required. Java 11+ is needed for Code Analyzer engines (PMD, SFGE).
• Data Cloud: Skills in the sf-datacloud-* family require an additional external sf data360 CLI plugin runtime.
• Links: Installation details are available at https://github.com/Jaganpro/sf-skills#installation.

Highlighted Details

• Extensive Skill Coverage: Encompasses Apex, Flow, LWC, SOQL, metadata, Data Cloud, integrations, testing, debugging, Agentforce, and DevOps.
• Automated Validation: Features hooks for Prettier, LSP compilation, custom scoring (up to 165 points for LWC), code analysis (PMD, ESLint), and debug log analysis.
• LSP Integration: Provides real-time syntax validation and auto-fix loops for Apex, LWC, and Agent Script.
• Live SOQL Query Plan Analysis: Integrates with Salesforce's explain API for immediate query optimization feedback.
• Specialized Agent Team: Seven AI agents are provided for roles like strategist, engineer, QA, and technical architect.

Maintenance & Community

This is a community-driven project primarily contributed to by Jag Valaiyapathy (CTA), with significant contributions from August Krys, Gnanasekaran Thoppae, David Ryan, and Shreyas Dhond. Volunteers are actively sought for testing across different CLIs.

Licensing & Compatibility

The project is released under the permissive MIT License, allowing for broad compatibility with commercial and closed-source projects.

Limitations & Caveats

The Data Cloud skills are in Beta/Community Preview and necessitate a separate CLI plugin installation. Several features, including specific cloud integrations (Sales, Service, Experience) and advanced AI capabilities (Einstein Copilot, Prediction Builder), are listed as "Planned" and not yet implemented. LSP support for Apex and Agent Script requires VS Code, while LWC LSP is standalone.