[WIMSE] IETF 127 Hackathon: WIMSE interop scenarios for agent call chains
Alexander PELOV <alexander.pelov@imt-atlantique.fr> Thu, 12 November 2026 17:55 UTC
Received: from zproxy2.enst.fr (zproxy2.enst.fr [IPv6:2a04:8ec0:0:2::1:221]) (using TLSv1.3 with cipher TLS_AES_256_GCM_SHA384 (256/256 bits) key-exchange x25519 server-signature ECDSA (prime256v1) server-digest SHA256) (No client certificate requested) by mx.ietf.org (Postfix) with ESMTPS id 15A1A46 for <wimse@ietf.org>; Thu, 12 Nov 2026 17:55:54 +0000 (UTC)
Authentication-Results: mx.ietf.org; dkim=pass header.d=imt-atlantique.fr header.s=50EA75E8-DE22-11E6-A6DE-0662BA474D24 header.b=iqHuIRl8; spf=pass (mx.ietf.org: domain of alexander.pelov@imt-atlantique.fr designates 2a04:8ec0:0:2::1:221 as permitted sender) smtp.mailfrom=alexander.pelov@imt-atlantique.fr; dmarc=pass (policy=none) header.from=imt-atlantique.fr
Received: from localhost (localhost [IPv6:::1]) by zproxy2.enst.fr (Postfix) with ESMTP id 42087807D9 for <wimse@ietf.org>; Thu, 12 Nov 2026 18:55:47 +0100 (CET)
Received: from zproxy2.enst.fr ([IPv6:::1]) by localhost (zproxy2.enst.fr [IPv6:::1]) (amavis, port 10032) with ESMTP id lDf9HQ_KYKTd for <wimse@ietf.org>; Thu, 12 Nov 2026 18:55:47 +0100 (CET)
Received: from localhost (localhost [IPv6:::1]) by zproxy2.enst.fr (Postfix) with ESMTP id 0A7348088B for <wimse@ietf.org>; Thu, 12 Nov 2026 18:55:47 +0100 (CET)
DKIM-Filter: OpenDKIM Filter v2.10.3 zproxy2.enst.fr 0A7348088B
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=imt-atlantique.fr; s=50EA75E8-DE22-11E6-A6DE-0662BA474D24; t=1794506147; bh=FGjDkjQBjlow8naE5OuKK1VMhQpRHXfO1DWPXRatHaI=; h=From:Mime-Version:Message-Id:Date:To; b=iqHuIRl8hxeClSgXJDkWpqdBnViZj7IG4hcAZBrlHSjIcmI/kMp4HYx/8gmWolnPJ 1G+qn9NEyWv0lHFqPOIPcrDXtNB/hObZJ2t9R1p16Yx6sG/3BCRmeIjDZkCiSEDeVi VFI3lS5gJmTmocvA8T3DrVSo0TA5t5T4js3RJJQU=
X-Virus-Scanned: amavis at
Received: from zproxy2.enst.fr ([IPv6:::1]) by localhost (zproxy2.enst.fr [IPv6:::1]) (amavis, port 10026) with ESMTP id HR6UYK12oIpb for <wimse@ietf.org>; Thu, 12 Nov 2026 18:55:46 +0100 (CET)
Received: from smtpclient.apple (unknown [212.133.41.107]) by zproxy2.enst.fr (Postfix) with ESMTPSA id 36676807D9 for <wimse@ietf.org>; Thu, 12 Nov 2026 18:55:45 +0100 (CET)
From: Alexander PELOV <alexander.pelov@imt-atlantique.fr>
Content-Type: multipart/alternative; boundary="Apple-Mail=_D0EEDEA3-254A-4876-B8DB-1A8F94FACFD2"
Mime-Version: 1.0 (Mac OS X Mail 16.0 \(3901.100.1.1.11\))
Message-Id: <1318AD2A-19D0-45A8-926D-8957D6826F2C@imt-atlantique.fr>
Date: Thu, 12 Nov 2026 18:54:28 +0100
To: wimse@ietf.org
X-Mailer: Apple Mail (2.3901.100.1.1.11)
X-Spamd-Bar: -----------------
Message-ID-Hash: KE423DSUTGISYXIFOIVBH2DTVDCVKIGK
X-Message-ID-Hash: KE423DSUTGISYXIFOIVBH2DTVDCVKIGK
X-MailFrom: alexander.pelov@imt-atlantique.fr
X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header
X-Mailman-Version: 3.3.10
Precedence: list
Subject: [WIMSE] IETF 127 Hackathon: WIMSE interop scenarios for agent call chains
List-Id: WIMSE Workload Identity in Multi-Service Environment <wimse.ietf.org>
Archived-At: <https://mailarchive.ietf.org/arch/msg/wimse/6ofJ6o3sK9H_XEPCQWn_7SVEza4>
List-Archive: <https://mailarchive.ietf.org/arch/browse/wimse>
List-Help: <mailto:wimse-request@ietf.org?subject=help>
List-Owner: <mailto:wimse-owner@ietf.org>
List-Post: <mailto:wimse@ietf.org>
List-Subscribe: <mailto:wimse-join@ietf.org>
List-Unsubscribe: <mailto:wimse-leave@ietf.org>
Dear all, At the IETF 127 Hackathon in San Francisco I'll be running interop scenarios across several independent WIMSE implementations. I’ll be using the AI2II project of the Hackathon to run the interop ( AI2II: AI-Assisted Protocol Implementation & Interoperability ). The focus is chains of AI agents calling each other across two trust domains, in line with the agent sections of draft-ietf-wimse-arch-08 and draft-ietf-wimse-aims-01. We’ll try to show the use of technologies from the following drafts: workload-creds-03, wpt-02, http-signature-08, identifier-04, oauth-transaction-tokens-11 and oauth-identity-chaining-17. You can see the full scenario we’ll be testing at the end of this mail. If you have ideas for other scenarios that you’d like to see tested, or if you can suggest implementations to run agains, please don’t hesitate to drop me a line. And if you want to join - remotely or on-site - it would be perfect! I'll report the results, and any spec ambiguities that we may find (which always pop during such interop events). Best regards, Alexander Pelov PS: Implementations I’ll use during the hackathon: - AgentMaurice - SPIRE (main, with the experimental WIT-SVID support) as a WIT issuer - go-spiffe v2.7.0 (experimental WIT-SVID support) - Cofide's WIMSE workload-to-workload PoC (Go; WIT + HTTP signatures) - An independent open-source Python implementation of WIT, WPT, HTTP signatures and Txn-Tokens, to be published - The test vectors from issue #419 in the ietf-wg-wimse GitHub organization Pointers to other implementations, especially of WPT or Txn-Tokens, are welcome. PPS. Scenario: Normal path - A user request enters trust domain A, and a Txn-Token is issued for it (subject, scope, transaction context) - Agent-to-agent calls are authenticated with a WIT plus an HTTP signature or a WPT, and the Txn-Token is bound to each hop's proof - Each hop authorizes using both the caller's workload identity and the Txn-Token (delegated or autonomous call, scope) - Crossing to trust domain B uses identity chaining (Txn-Token, then JWT grant, then access token); the Txn-Token itself never leaves its domain - Trust anchors are exchanged out of band only, with no key discovery from tokens - Every hop logs the same transaction ID, and the logs and packet captures merge into one end-to-end audit trail Negative cases - WIT presented without proof of possession - Replayed request, and request sent to the wrong audience - Txn-Token swapped after signing, or taken from another trust domain - Prompt injection that asks an agent to read a resource ou - Prompt injection that asks an agent to reveal its credentials - Autonomous identity attempting a delegated action - Revoking an agent mid-transaction - Workload from an untrusted trust domain
- [WIMSE] IETF 127 Hackathon: WIMSE interop scenari… Alexander PELOV
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Iman Schrock
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Ben Le
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Iman Schrock
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Marc Miro i Rierola
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Ben Le
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Arun Thallapelly
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Ben Le
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Kieran Sweeney
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Jacob Stewart
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Arun Thallapelly
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Alexander PELOV
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Arun Thallapelly
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… Kieran Sweeney
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… ben.le
- [WIMSE] Re: IETF 127 Hackathon: WIMSE interop sce… ben.le