What should an AI agent's documentation make explicit?

Three questions shaped our Agent Manifest Kit:

• What can the agent do, and what is outside its scope?

• Which compatibility claims have actually been verified?

• What access does a user grant before trying it?

We organized the answers into six Framer pages: Home, Quickstart, Skills, Compatibility, Security, and Changelog. One lesson from refining the navigation: each page needs a clear next step, especially between Compatibility and Security.

The starter is free to remix. It provides the documentation website; it does not include a hosted agent, live discovery files, or official A2A conformance. Replace the demo content with your own verified details before publishing.

Disclosure: we created this template and may earn a commission from qualifying paid Framer subscriptions associated with its referral flow.

Which scope boundary do you make explicit before publishing an AI tool?