Cryptographic Separation of Untrusted Data-Plane Context from Control-Plane Tool Invocations in Autonomous AI Agents

Status of This Memo

This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶

Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶

Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶

This Internet-Draft will expire on 2 May 2027.¶

Copyright Notice

Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved.¶

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶

Table of Contents

1. Introduction and Problem Statement

In classical microservices, control-plane requests are triggered by deterministic code paths. In autonomous AI agents, however, a single LLM inference window blends trusted system instructions with untrusted data-plane inputs (e.g., third-party documents, webhooks, or unverified peer agent outputs). Neither OAuth 2.0 Token Exchange (RFC 8693) nor WIMSE Workload Identity Tokens carry a cryptographic commitment to the taint lattice of the context window active at the moment of tool selection.¶

This specification complements the foundational execution-trace verification architecture defined in draft-durga-agenttrust-trace-verification-00 by introducing deterministic, fail-closed wire semantics specifically addressing Agentic Context Taint Isolation.¶

2. Conventions and Definitions

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in BCP 14 ([RFC2119] [RFC8174]) when, and only when, they appear in all capitals, as shown here.¶

3. Context Taint Attestation (CTA) Envelope and HTTP/MCP Header Binding

Every control-plane tool invocation emitted by a conforming agent runtime MUST include an Agent-Ctx-Isolation header (or _meta.ctxIsolation object in JSON-RPC/MCP) carrying a signed Context Taint Attestation object serialized using RFC 8785 JCS or RFC 8949 Deterministic CBOR:¶

{
  "cta_version": "1.0",
  "invocation_id": "urn:uuid:7f9c2b10-44a1-49e2-910b-88a12c34e901",
  "parent_trace_ref": "sha256:9f86d081884c7d659a2feaa0c55ad015a3bf4f1b2b0b822cd15d6c15b0f00a08",
  "system_prompt_digest": "sha256:4b227777d4dd1fc61c6f884f48641d02b4d121d3fd328cb08b5531fcacdabf8a",
  "data_plane_merkle_root": "sha256:e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855",
  "taint_lattice_level": "UNTRUSTED_INGESTED",
  "tool_effect_class": "STATE_MUTATING",
  "step_up_receipt_hash": "sha256:d4735e3a265e16eee03f59718b9b5d03019c07d8b6c51f90da3a666eec13ab35",
  "iat": 1793404800,
  "exp": 1793404860,
  "signature": {
    "alg": "Ed25519",
    "kid": "sha256:1b4f0e9851971998e732078544c96b36c3d01cedf7caa332359d6f1d83567014",
    "sig": "base64url(Ed25519(JCS(payload)))"
  }
}

4. Normative Fail-Closed Verification Rules

Conforming verifiers and relying-party gateways MUST enforce the following deterministic invariants before releasing any downstream side effect:¶

1. TAINT_LATTICE_MONOTONICITY: Once an inference context ingests any byte from an unverified external origin, the runtime MUST transition taint_lattice_level from TRUSTED_ONLY to MIXED_TAINT or UNTRUSTED_INGESTED for the remainder of that context session unless a cryptographic context-reset boundary is executed.¶
2. CONFUSED_DEPUTY_FAIL_CLOSED: A relying party receiving a tool invocation with tool_effect_class set to STATE_MUTATING or PRIVILEGED_ACTUATE and taint_lattice_level != TRUSTED_ONLY MUST reject the invocation with error code ERR_TAINTED_CONTROL_PLANE unless a valid, unexpired step_up_receipt_hash signed by an independent policy decision engine or human verifier is attached.¶
3. CONTEXT_MERKLE_BINDING: The data_plane_merkle_root MUST be computed over the ordered RFC 8785 JCS-canonicalized chunks of all retrieved documents and tool outputs present in the active context window.¶

5. Security Considerations

All pre-image objects MUST be canonicalized using RFC 8785 JSON Canonicalization Scheme (JCS) prior to signature generation or verification under RFC 8032 Ed25519. Any clock skew exceeding 60 seconds, digest mismatch, or missing cryptographic binding MUST cause the verifier to fail closed and emit an auditable security rejection receipt.¶

6. IANA Considerations

This document requests IANA registration of the media type application/agent-ctx-isolation+json and the registry IETF Agent Context Taint Lattice Levels (TRUSTED_ONLY, MIXED_TAINT, UNTRUSTED_INGESTED, QUARANTINED), alongside the following header fields:¶

• HTTP/MCP Header Field: Agent-Ctx-Isolation¶
• HTTP/MCP Header Field: Agent-Step-Up-Receipt¶

Author's Address