Hi Steve, Thank you very much for your feedback on our draft, draft-chen-oauth-agent-intent-token-00. Your comments are incredibly valuable, and we appreciate you taking the time to review our work. We particularly appreciate you highlighting the challenge of "divergence that builds up across requests." The concept of cumulative intent drift is a critical security aspect that deserves explicit attention, and your framing of the problem is spot-on. To ensure we properly address this, we have logged it as an issue on GitHub for tracking and further discussion. Based on your excellent suggestion, we are planning to make the following additions in the next revision of the draft: A new subsection within "Security Considerations" dedicated to discussing the risks of Cumulative Intent Drift and outlining mitigation strategies, such as stateful analysis by the Authorization Server. Strengthening the definition and role of intent_id as a key mechanism to enable this sequential analysis across multiple token exchanges. Adding a "Related Work" section to acknowledge AuthR. We agree it is highly relevant, and we appreciate you bringing it to our attention. We would be delighted to take you up on your offer to compare notes as we develop these sections. To foster a broader conversation on this important topic, we've copied the OAuth Working Group mailing list. We believe the entire group can benefit from this discussion. Best, Meiling
[email protected] From: Steve Tout Date: 2026-10-24 11:55 To: draft-chen-oauth-agent-intent-token.authors Subject: Congratulations on the Intent Token draft Dear Meiling, Li, and Peter, Congratulations on publishing draft-chen-oauth-agent-intent-token. Treating intent as a signed, first-class object that the authorization server can evaluate against is an important step for agent authorization, and I’m glad to see it in front of the OAuth community. I started working on this problem earlier this year, which led me to publish the first draft of the AuthR (Authorship Representation) spec as open source in April: https://github.com/Identient/authr Intent is one of AuthR’s core primitives, but the spec also treats drift as essential to how it operates. Each request can match the original intent on its own while the sequence of actions diverges from it. That makes detecting drift across a chain of requests and delegations as important as recording intent at the start. AuthR has been implemented in the NeuroNest agentic IDE. As you develop the Security Considerations section, I’d be curious how you’re thinking about divergence that builds up across requests rather than within a single one. AuthR may be useful there as related work, and I’d be happy to compare notes anytime. Regards, Steve -- Best, Steve Tout Founder & CEO, Identient Decision Intelligence Built on Verifiable AI and Accountable Expertise [email protected] | https://www.identient.com
_______________________________________________ OAuth mailing list -- [email protected] To unsubscribe send an email to [email protected]
