Best practices & resources
Things to remember as you start creating your own instructions.
You've learned how skills and rules work, seen how to create and manage them in Webflow, and watched an agent put them to use. Now, let's review some best practices as you get started with your own instructions.
Before, during, and after using Agent Instructions
Before you start
- Start with one repeated problem or decision you want the agent to handle better. That's usually a good candidate for an instruction.
- If your site already has components and variables, reference them directly instead of describing them in text.
- Not sure where to begin? Generate a skill from your site, then add any team context or decisions the site can't show on its own.
While you're working
- At the start of a relevant session, ask the agent to read your Agent Instructions: "Before we start, please read my Agent Instructions for this site."
- When the agent makes a choice you wouldn't have made, treat it as feedback. If it keeps happening, add or refine an instruction.
- Make those updates while the context is still fresh.
After a build or project handoff
- For a single handoff, export your instructions as a ZIP file for your teammate or client.
- Managing several sites for the same team or client? Consider moving shared guidance into a Shared Library.
- Update your instructions when your design system, content structure, or brand standards change. Stale guidance can mislead an agent just like outdated documentation can mislead a new teammate.
Additional resources
- Create and manage AI agent instructions — the Webflow Help Center reference for skills, rules, references, and the Instructions panel
- Agent Instructions — Webflow MCP documentation — Agent Instructions in the context of the Webflow MCP server
- Agent skills: What, why, and Webflow — how agent skills work and why they matter for MCP workflows
- webflow/webflow-skills — Webflow-built skill files you can reference or use directly
- Get started with Webflow MCP — recommended Webflow University starting point for connecting and working with Webflow MCP
- AI guardrails for teams — a Webflow University course covering workspace AI settings, roles, and governance for teams
That's a wrap
Click the Complete course button in the progress box on the right, then head back to the Webflow University Courses page to start your next course.