Agent checksum
agent_checksum · agent checksums
A Secure Hash Algorithm 256 (SHA-256) fingerprint of an AI agent's system prompt, tool definitions and model settings, which an identity server can compare with a registered value before issuing a token.
In our briefings
The claim each of the newest briefings on Agent checksum would have you leave with.
Revision -02 adds a section on what a server holding the credentials cannot promise. Compromise of the Delegation Server (DS), its operator or the path that releases the keys protecting stored credentials lets an attacker use those credentials with no Delegation Token at all, and Sweeney writes that "Keeping credentials away from Agents does not remove this exposure; it concentrates it at the DS." The server's signed receipts prove it authorized an action under its own records, but, in his words, "it does not prove that the Subject, rather than the operator, approved the grant". Approval signed with a key the user holds is left to a later revision. The revocation section now states its limits too. The one-second and five-second targets stand, yet "revoked_count reports invalidation, not closure": a token the server exchanged for direct use at a resource server, or issued under the offline profile, can be honored by an enforcement point off the server's path until it expires, at most an hour after issue, unless that point checks authority or status. Synchronous cascading revocation holds within one Delegation Server; enforcing it across servers or domains is out of scope.
Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claimBI-FIND-2026-1939
The agent identity in this draft is a fingerprint the agent's own side computes and reports. A client library hashes the agent's identifier, its prompt, the names and descriptions of its tools and its model settings; the identity provider stores that value at registration and, when a token is requested, compares what the client says against what it stored. It never observes the running agent. By default the hash covers a tool's description and not its code, and the draft lists a compromised client library as a threat with no defence defined. Any change to prompt, tools or settings means 'the agent MUST re-register'. What is attested is a configuration as declared, not behaviour.
The 'intent' in the Agentic JWT draft is a workflow step, and the agent reports its own fingerprintBI-FIND-2026-0306
Goenka, Pathak and Asthana argue that the agent payment protocols of January 2026 let money move before anyone checked the agent's work. In their account AP2 "functions strictly as a messaging layer" that assumes the merchant is honest and has no way to check execution before funds are captured; Coinbase's x402 follows a pay-first model in which funds are released before the service is delivered; and the Agent Commerce Protocol leans on reputation scores, which new agents cannot have. No protocol they reviewed, they conclude, covers the whole path from delegation to settlement to dispute.
A January design paper argued AP2 and x402 move money before anyone checks the work, and proposed escrowBI-FIND-2026-1228
Cited in 4 briefings
Every briefing that discusses this term, newest first.
- Sweeney's delegation draft concedes its server is trusted with users' credentials, and drops an interop claim
Draft · · IETF Datatracker · Kieran Sweeney · Oct 13, 2026
- The 'intent' in the Agentic JWT draft is a workflow step, and the agent reports its own fingerprint
Draft · · IETF Datatracker · Abhishek Goswami · Jul 1, 2026
- A January design paper argued AP2 and x402 move money before anyone checks the work, and proposed escrow
Paper · · arXiv · Mehul Goenka et al. · Jan 30, 2026
- Agentic JWT binds an agent's token to its task and its own code, but it is one author's draft, not a standard
Paper · · arXiv (cs.CR preprint 2509.13597) · Abhishek Goswami · Sep 16, 2025