AI Assistant Workflow With This RAG

Overview

This guide defines how to use AI assistants against this repository so outputs stay grounded in your RAG content instead of generic Salesforce advice.

Canonical Source Priority

1. rag/ content pages.
2. rag/rag-index.md for discovery.
3. rag/rag-library.json for programmatic retrieval.
4. Knowledge/ and other docs as secondary reference material.

Assistant Prompting Pattern

When starting any assistant session, provide:

• repository purpose,
• source priority order,
• requirement to cite exact rag/ files,
• requirement to refuse unsupported claims.

Recommended Session Flow

1. Start with query intent.
2. Ask assistant to identify relevant rag/ files first.
3. Load only those files.
4. Produce evidence-first answer with citations.
5. If evidence is weak, return “insufficient evidence” and request more context.

Cursor Workflow

• Open repo root so rag/ is indexed.
• Use direct file mentions (@rag/...) for focused retrieval.
• Keep prompts domain-scoped (e.g., integrations, security, testing).

Chat Assistant Workflow (Claude/ChatGPT)

• Paste context contract and source priority.
• Include relevant file snippets or file list from rag/rag-index.md.
• Require citation paths in final answer.

Quality Guardrails

• No uncited architecture claims.
• No invented Salesforce limits or API behavior.
• No production recommendations without linked patterns in rag/.
• If conflicting guidance exists, call it out explicitly.

Sources Used

• docs/ai-products-guide.md - Detailed operating instructions for Cursor/Claude/ChatGPT with this repository

Related Patterns

• Source Ingestion Status
• RAG Style Guide
• User Readiness for Salesforce