Best Practices

Here’s a list of best practices for using Agentforce Vibes Extension

• Be specific about your goals: Instead of “help with testing,” say “run all Apex tests in my AccountController class”.
• Provide context: Mention specific orgs, projects, or files you’re working with.
• Use natural language: Describe what you want to accomplish rather than trying to format commands.
• Ask for explanations: Request explanations of actions for learning purposes.

• Review before approving: Always check proposed actions before confirming execution.
• Use incremental steps: Break complex workflows into manageable pieces.
• Monitor progress: Pay attention to progress indicators and error messages.
• Verify results: Check that completed actions meet your expectations.

• Verify org contexts: Ensure actions target the correct Salesforce org.
• Review metadata changes: Check what’s being deployed before confirmation.
• Use appropriate environments: Use sandboxes for testing, production org for final deployment.
• Monitor permissions: Ensure you have appropriate access for requested actions.

• Use sample projects: Start with dreamhouse-lwc or ebikes-lwc for rich testing.
• Test incrementally: Verify each step before proceeding to the next.
• Document your setup: Note any custom configurations for team sharing.
• Regular updates: Keep the VS Code extension updated as new versions are released.

• Use sandbox orgs: Connect to sandbox environments for testing Agentforce.
• Review permissions: Ensure your user has appropriate development permissions.
• Monitor org access: Regularly review and refresh org authorizations.
• Backup important work: Maintain version control for critical development.

• Match the model to the task: Use a balanced model (Sonnet) for everyday questions, explanations, and single-file edits. Switch to the most capable model (Opus) for complex multi-file builds, architectural planning, and Plan Mode approval.
• Switch mid-session when scope changes: You can change models without losing context. Start on a faster model while exploring, then switch up before generating a plan or executing a complex task.
• Monitor token usage: The context-window indicator shows your current usage. More capable models consume more tokens per response. Start a fresh conversation rather than continuing in a heavily compacted session.
• Factor in Flex Credits: If your org uses metered billing, model choice affects cost. Use balanced or fast models for exploration and iteration. Reserve premium models for tasks where output quality justifies the expense. See Usage and Billing.

• Standardize configurations: Share common settings across your development team.
• Document custom tools: If you add custom MCP servers, document them for the team.
• Regular sync: Ensure all team members use compatible extension versions.