Execution Context Tokens for Distributed Agentic Workflows

1.1. Motivation

The Workload Identity in Multi System Environments (WIMSE) framework [I-D.ietf-wimse-arch] provides robust workload authentication through Workload Identity Tokens (WIT) and Workload Proof Tokens (WPT). The WIMSE service-to-service protocol [I-D.ietf-wimse-s2s-protocol] defines how workloads authenticate each other across call chains using the Workload-Identity and Workload-Proof-Token HTTP headers.¶

However, workload identity alone does not address execution accountability. Knowing who performed an action does not prove what was done, what policy was applied, or whether compliance requirements were satisfied at each decision point.¶

Regulated environments increasingly deploy autonomous agents that coordinate across organizational boundaries. Multiple regulatory frameworks motivate the need for structured execution records:¶

• The EU Artificial Intelligence Act [EU-AI-ACT] Article 12 requires high-risk AI systems to be designed with capabilities enabling automatic recording of events ("logs") while the system is operating.¶

• The U.S. FDA 21 CFR Part 11 [FDA-21CFR11] requires computer-generated, timestamped audit trails that independently record the date, time, operator identity, and actions taken (Section 11.10(e)).¶

• The Markets in Financial Instruments Directive (MiFID II) [MIFID-II] requires firms to maintain records of transactions and orders that are sufficient to enable supervisory authorities to monitor compliance.¶

• The Digital Operational Resilience Act (DORA) [DORA] Article 12 requires financial entities to have logging policies that record ICT activities and anomalies.¶

This document defines an extension to the WIMSE architecture that addresses the gap between workload identity and execution accountability. WIMSE authenticates agents; this extension records what they did, in what order, and what policy was evaluated.¶

As identified in [I-D.ni-wimse-ai-agent-identity], call context in agentic workflows must always be visible and preserved. ECTs provide a mechanism to address this requirement with cryptographic assurances.¶

1.2. Problem Statement

Three core gaps exist in current approaches to regulated agentic systems:¶

1. WIMSE authenticates agents but does not record what they actually did. A WIT proves "Agent A is authorized" but not "Agent A executed Task X, under Policy Y, producing Output Z."¶

2. No standard mechanism exists to record policy evaluation outcomes at each decision point in a multi-agent workflow.¶

3. No mechanism exists to cryptographically link compensation and rollback decisions to original actions.¶

Existing observability tools such as distributed tracing [OPENTELEMETRY] provide visibility for debugging and monitoring but do not provide cryptographic assurances. Tracing data is not cryptographically signed, not tamper-evident, and not designed for regulatory audit scenarios.¶

1.3. Scope and Applicability

This document defines:¶

• The Execution Context Token (ECT) format (Section 4)¶

• DAG structure for task dependency ordering (Section 6)¶

• Policy checkpoint recording (Section 4.2.3)¶

• Integration with the WIMSE identity framework (Section 3)¶

• An HTTP header for ECT transport (Section 5)¶

• Audit ledger interface requirements (Section 8)¶

The following are out of scope and are handled by WIMSE:¶

• Workload authentication and identity provisioning¶

• Key distribution and management¶

• Trust domain establishment and management¶

• Credential lifecycle management¶

1.4. Relationship to Regulatory Compliance

ECTs are a technical mechanism that can support compliance programs by providing structured, cryptographically signed execution records. ECTs do not by themselves constitute compliance with any regulatory framework referenced in this document.¶

Compliance with each referenced regulation requires organizational controls, policies, procedures, validation, and governance measures beyond the scope of this specification. The regulatory references in this document are intended to motivate the design requirements, not to claim that implementing ECTs satisfies these regulations.¶

ECTs provide evidence of claimed execution ordering and policy evaluation. They do not independently verify that the claimed execution actually occurred as described, that the policy evaluation was correct, or that the agent faithfully performed the stated action. The trustworthiness of ECT claims depends on the trustworthiness of the signing agent and the integrity of the broader deployment environment.¶

3.1. WIMSE Foundation

The WIMSE architecture [I-D.ietf-wimse-arch] defines:¶

• Workload Identity Tokens (WIT) that prove a workload's identity within a trust domain ("I am Agent X in trust domain Y")¶

• Workload Proof Tokens (WPT) that prove possession of the private key associated with a WIT ("I control the key for Agent X")¶

• Multi-hop authentication via the service-to-service protocol [I-D.ietf-wimse-s2s-protocol]¶

The following execution accountability needs are complementary to the WIMSE scope and are not addressed by workload identity alone:¶

• Recording what agents actually do with their authenticated identity¶

• Recording policy evaluation outcomes at each hop¶

• Maintaining structured execution records¶

• Linking compensation or rollback actions to original tasks¶

3.2. Extension Model

ECTs extend WIMSE by adding an execution accountability layer between the identity layer and the application layer:¶

+--------------------------------------------------+
|  WIMSE Layer (Identity)                          |
|    WIT: "I am Agent X (spiffe://td/agent/x)"    |
|    WPT: "I prove I control the key for Agent X"  |
+--------------------------------------------------+
                       |
                       v
+--------------------------------------------------+
|  ECT Layer (Execution Accountability) [This Spec]|
|    ECT: "Task executed, dependencies met,        |
|          policy evaluated, outcome recorded"      |
+--------------------------------------------------+
                       |
                       v
+--------------------------------------------------+
|  Ledger Layer (Immutable Record)                 |
|    "All ECTs appended to audit ledger"           |
+--------------------------------------------------+

This extension reuses the WIMSE signing model, extends JWT claims using standard JWT extensibility [RFC7519], and maintains WIMSE concepts including trust domains and workload identifiers.¶

3.3. Integration Points

An ECT integrates with the WIMSE identity framework through the following mechanisms:¶

• The ECT JOSE header "kid" parameter MUST reference the public key identifier from the agent's WIT.¶

• The ECT "iss" claim MUST use the WIMSE workload identifier format (a SPIFFE ID [SPIFFE]).¶

• The ECT MUST be signed with the same private key used to generate the agent's WPT.¶

• The ECT signing algorithm (JOSE header "alg" parameter) MUST match the algorithm used in the corresponding WIT.¶

When an agent makes an HTTP request to another agent, the three tokens are carried in their respective HTTP header fields:¶

HTTP Request from Agent A to Agent B:
  Workload-Identity:    <WIT for Agent A>
  Workload-Proof-Token: <WPT proving A controls key>
  Execution-Context:    <ECT recording what A did>

The receiving agent (Agent B) verifies in order:¶

1. WIT and WPT (WIMSE layer): Proves who Agent A is and that the request is authentic.¶

2. ECT (this extension): Records what Agent A did, what policy was evaluated, and what precedent tasks exist.¶

3. Ledger: Appends the verified ECT to the audit ledger.¶

4.1. JOSE Header

The ECT JOSE header MUST contain the following parameters:¶

{
  "alg": "ES256",
  "typ": "wimse-exec+jwt",
  "kid": "agent-a-key-id-123"
}

alg:

REQUIRED. The digital signature algorithm used to sign the ECT. MUST match the algorithm in the corresponding WIT. Implementations MUST support ES256 [RFC7518]. The "alg" value MUST NOT be "none". Symmetric algorithms (e.g., HS256, HS384, HS512) MUST NOT be used, as ECTs require asymmetric signatures for non-repudiation.¶

typ:

REQUIRED. MUST be set to "wimse-exec+jwt" to distinguish ECTs from other JWT types, consistent with the WIMSE convention for type parameter values.¶

kid:

REQUIRED. The key identifier referencing the public key from the agent's WIT [RFC7517]. Used by verifiers to look up the correct public key for signature verification.¶

4.2. JWT Claims

The ECT payload contains both WIMSE-compatible standard JWT claims and execution context claims defined by this specification.¶

4.3. Complete ECT Example

The following is a complete ECT payload example:¶

{
  "iss": "spiffe://example.com/agent/clinical",
  "sub": "spiffe://example.com/agent/clinical",
  "aud": "spiffe://example.com/agent/safety",
  "iat": 1772064150,
  "exp": 1772064750,

  "wid": "a0b1c2d3-e4f5-6789-abcd-ef0123456789",
  "tid": "550e8400-e29b-41d4-a716-446655440001",
  "exec_act": "recommend_treatment",
  "par": [],

  "pol": "clinical_reasoning_policy_v2",
  "pol_decision": "approved",
  "pol_enforcer": "spiffe://example.com/policy/clinical-engine",
  "pol_timestamp": 1772064145,

  "inp_hash": "sha-256:n4bQgYhMfWWaL-qgxVrQFaO_TxsrC4Is0V1sFbDwCgg",
  "out_hash": "sha-256:LCa0a2j_xo_5m0U8HTBBNBNCLXBkg7-g-YpeiGJm564",
  "inp_classification": "confidential",
  "exec_time_ms": 245,
  "regulated_domain": "medtech",
  "model_version": "clinical-reasoning-v4.2",

  "witnessed_by": [
    "spiffe://example.com/audit/observer-1"
  ]
}

5.1. Execution-Context Header Field

This specification defines the Execution-Context HTTP header field [RFC9110] for transporting ECTs between agents.¶

The header field value is the ECT in JWS Compact Serialization format [RFC7515]. The value consists of three Base64url-encoded parts separated by period (".") characters.¶

An agent sending a request to another agent includes the Execution-Context header alongside the WIMSE Workload-Identity and Workload-Proof-Token headers:¶

GET /api/safety-check HTTP/1.1
Host: safety-agent.example.com
Workload-Identity: eyJhbGci...WIT...
Workload-Proof-Token: eyJhbGci...WPT...
Execution-Context: eyJhbGci...ECT...

When multiple parent tasks contribute context to a single request, multiple Execution-Context header field lines MAY be included, each carrying a separate ECT in JWS Compact Serialization format.¶

6.1. Overview

ECTs form a Directed Acyclic Graph (DAG) where each task references its parent tasks via the "par" claim. This structure provides a cryptographically signed record of execution ordering, enabling auditors to reconstruct the complete workflow and verify that required predecessor tasks were recorded before dependent tasks.¶

6.2. Validation Rules

When receiving and verifying an ECT, implementations MUST perform the following DAG validation steps:¶

1. Task ID Uniqueness: The "tid" claim MUST be unique within the applicable scope (the workflow identified by "wid", or the entire ledger if "wid" is absent). If a task with the same "tid" already exists, the ECT MUST be rejected.¶

2. Parent Existence: Every task identifier listed in the "par" array MUST correspond to a task that has been previously recorded in the ledger. If any parent task is not found, the ECT MUST be rejected.¶

3. Temporal Ordering: The "iat" value of every parent task MUST be less than the "iat" value of the current task plus a configurable clock skew tolerance (RECOMMENDED: 30 seconds). If any parent task has an "iat" that violates this constraint, the ECT MUST be rejected.¶

4. Acyclicity: Following the chain of parent references MUST NOT lead back to the current task's "tid". If a cycle is detected, the ECT MUST be rejected.¶

5. Trust Domain Consistency: Parent tasks SHOULD belong to the same trust domain or to a trust domain with which a federation relationship has been established.¶

6.3. DAG Validation Algorithm

The following pseudocode describes the DAG validation procedure:¶

function validate_dag(ect, ledger, clock_skew_tolerance):
  // Step 1: Uniqueness check
  if ledger.contains(ect.tid, ect.wid):
    return error("Task ID already exists in ledger")

  // Step 2: Parent existence and temporal ordering
  for parent_id in ect.par:
    parent = ledger.get(parent_id)
    if parent is null:
      return error("Parent task not found: " + parent_id)
    if parent.iat >= ect.iat + clock_skew_tolerance:
      return error("Parent task not earlier than current")

  // Step 3: Cycle detection
  visited = set()
  if has_cycle(ect.tid, ect.par, ledger, visited):
    return error("Circular dependency detected")

  return success

function has_cycle(target_tid, parent_ids, ledger, visited):
  for parent_id in parent_ids:
    if parent_id == target_tid:
      return true
    if parent_id in visited:
      continue
    visited.add(parent_id)
    parent = ledger.get(parent_id)
    if parent is not null:
      if has_cycle(target_tid, parent.par, ledger, visited):
        return true
  return false

The cycle detection traverses the ancestor graph rooted at the current task's parents. The complexity is O(V) where V is the number of ancestor nodes reachable from the current task's parent references. For typical workflows with shallow DAGs, this is efficient. Implementations SHOULD cache cycle detection results for previously verified tasks to avoid redundant traversals.¶

7.1. Verification Procedure

When an agent receives an ECT, it MUST perform the following verification steps in order:¶

1. Parse the JWS Compact Serialization to extract the JOSE header, payload, and signature components per [RFC7515].¶

2. Verify that the "typ" header parameter is "wimse-exec+jwt".¶

3. Verify that the "alg" header parameter is not "none" and is not a symmetric algorithm.¶

4. Verify the "kid" header parameter references a known, valid public key from a WIT within the trust domain.¶

5. Retrieve the public key identified by "kid" and verify the JWS signature per [RFC7515] Section 5.2.¶

6. Verify the "alg" header parameter matches the algorithm in the corresponding WIT.¶

7. Verify the "iss" claim matches the "sub" claim of the WIT associated with the "kid" public key.¶

8. Verify the "aud" claim contains the verifier's own workload identity or an expected recipient identifier.¶

9. Verify the "exp" claim indicates the ECT has not expired.¶

10. Verify the "iat" claim is not unreasonably far in the past (implementation-specific threshold, RECOMMENDED maximum of 15 minutes).¶

11. Verify all required claims ("tid", "exec_act", "par", "pol", "pol_decision") are present and well-formed.¶

12. Verify "pol_decision" is one of "approved", "rejected", or "pending_human_review".¶

13. Perform DAG validation per Section 6.¶

14. If all checks pass, the ECT MUST be appended to the audit ledger.¶

If any verification step fails, the ECT MUST be rejected and the failure MUST be logged for audit purposes. Error messages SHOULD NOT reveal whether specific parent task IDs exist in the ledger, to prevent information disclosure.¶

7.2. Verification Pseudocode

function verify_ect(ect_jws, verifier_id,
                     trust_domain_keys, ledger):
  // Parse JWS
  (header, payload, signature) = parse_jws(ect_jws)

  // Verify header
  if header.typ != "wimse-exec+jwt":
    return reject("Invalid typ parameter")

  if header.alg == "none" or is_symmetric(header.alg):
    return reject("Prohibited algorithm")

  // Look up public key
  public_key = trust_domain_keys.get(header.kid)
  if public_key is null:
    return reject("Unknown key identifier")

  // Verify signature
  if not verify_jws_signature(header, payload,
                               signature, public_key):
    return reject("Invalid signature")

  // Verify algorithm alignment
  wit = get_wit_for_key(header.kid)
  if header.alg != wit.alg:
    return reject("Algorithm mismatch with WIT")

  // Verify issuer matches WIT subject
  if payload.iss != wit.sub:
    return reject("Issuer does not match WIT subject")

  // Verify audience
  if verifier_id not in payload.aud:
    return reject("ECT not intended for this recipient")

  // Verify not expired
  if payload.exp < current_time():
    return reject("ECT has expired")

  // Verify iat freshness
  if payload.iat < current_time() - max_age_threshold:
    return reject("ECT issued too long ago")

  // Verify required claims
  for claim in ["tid", "exec_act", "par",
                 "pol", "pol_decision"]:
    if claim not in payload:
      return reject("Missing required claim: " + claim)

  // Validate pol_decision value
  if payload.pol_decision not in
     ["approved", "rejected", "pending_human_review"]:
    return reject("Invalid pol_decision value")

  // Validate DAG
  result = validate_dag(payload, ledger,
                         clock_skew_tolerance)
  if result is error:
    return reject("DAG validation failed")

  // All checks passed; append to ledger
  ledger.append(payload)
  return accept

8.1. Overview

ECTs are designed to be recorded in an immutable audit ledger for compliance verification and post-hoc analysis. This specification defines the logical interface for the ledger but does not mandate a specific storage technology. Implementations MAY use append-only logs, databases with cryptographic commitment schemes, distributed ledgers, or any storage mechanism that provides the required properties.¶

8.2. Required Properties

An audit ledger implementation MUST provide:¶

1. Append-only semantics: Once an ECT is recorded, it MUST NOT be modified or deleted.¶

2. Ordering: The ledger MUST maintain a total ordering of ECT entries via a monotonically increasing sequence number.¶

3. Lookup by task ID: The ledger MUST support efficient retrieval of ECT entries by "tid" value.¶

4. Integrity verification: The ledger SHOULD provide a mechanism to verify that no entries have been tampered with (e.g., hash chains or Merkle trees).¶

The ledger SHOULD be maintained by an entity independent of the workflow agents to reduce the risk of collusion.¶

8.3. Ledger Entry Structure

Each ledger entry is a logical record containing:¶

{
  "ledger_sequence": 42,
  "task_id": "550e8400-e29b-41d4-a716-446655440001",
  "agent_id": "spiffe://example.com/agent/clinical",
  "action": "recommend_treatment",
  "parents": [],
  "ect_jws": "eyJhbGciOiJFUzI1NiIs...<complete JWS>",
  "signature_verified": true,
  "verification_timestamp": "2026-02-24T15:42:31.000Z",
  "stored_timestamp": "2026-02-24T15:42:31.050Z"
}

The "ect_jws" field contains the full JWS Compact Serialization and is the authoritative record. The other fields ("agent_id", "action", "parents") are convenience indexes derived from the ECT payload; if they disagree with the JWS payload, the JWS payload takes precedence.¶

9.1. Medical Device SDLC Workflow

In a medical device software development lifecycle (SDLC), AI agents assist across multiple phases from requirements analysis through release approval. Regulatory frameworks including [FDA-21CFR11] Section 11.10(e) and [EU-MDR] require audit trails documenting the complete development process for software used in medical devices.¶

Agent A (Spec Reviewer):
  tid: task-001    par: []
  exec_act: review_requirements_spec
  pol: spec_review_policy_v2     pol_decision: approved

Agent B (Code Generator):
  tid: task-002    par: [task-001]
  exec_act: implement_module
  pol: coding_standards_v3       pol_decision: approved

Agent C (Test Agent):
  tid: task-003    par: [task-002]
  exec_act: execute_test_suite
  pol: test_coverage_policy_v1   pol_decision: approved

Agent D (Build Agent):
  tid: task-004    par: [task-003]
  exec_act: build_release_artifact
  pol: build_validation_v2       pol_decision: approved

Human Release Manager:
  tid: task-005    par: [task-004]
  exec_act: approve_release
  pol: release_approval_policy   pol_decision: approved
  pol_enforcer: spiffe://meddev.example/human/release-mgr-42
  witnessed_by: [spiffe://meddev.example/audit/qa-observer-1]

ECTs record that requirements were reviewed before implementation began, that tests were executed against the implemented code, that the build artifact was validated, and that a human release manager explicitly approved the release. The DAG structure ensures no phase was skipped or reordered.¶

task-001 (review_requirements_spec)
  |
  v
task-002 (implement_module)
  |
  v
task-003 (execute_test_suite)
  |
  v
task-004 (build_release_artifact)
  |
  v
task-005 (approve_release) [human, witnessed]

9.2. Financial Trading Workflow

In a financial trading workflow, agents perform risk assessment, compliance verification, and trade execution. The DAG structure records that compliance checks were evaluated before trade execution.¶

Agent A (Risk Assessment):
  tid: task-001    par: []
  exec_act: calculate_risk_exposure
  pol: risk_limits_policy_v2  pol_decision: approved

Agent B (Compliance):
  tid: task-002    par: [task-001]
  exec_act: verify_compliance
  pol: compliance_check_v1    pol_decision: approved

Agent C (Execution):
  tid: task-003    par: [task-002]
  exec_act: execute_trade
  pol: execution_policy_v3    pol_decision: approved

This can contribute to compliance with:¶

• [MIFID-II]: ECTs provide cryptographic records of the execution sequence that can support transaction audit requirements.¶

• [DORA] Article 12: ECTs contribute to ICT activity logging.¶

• [EU-AI-ACT] Article 12: Logging of decisions made by AI-driven systems.¶

9.3. Compensation and Rollback

When a compliance violation is discovered after execution, ECTs provide a mechanism to record authorized compensation actions with a cryptographic link to the original task:¶

{
  "iss": "spiffe://bank.com/agent/operations",
  "sub": "spiffe://bank.com/agent/operations",
  "aud": "spiffe://bank.com/system/ledger",
  "iat": 1772150550,
  "exp": 1772151150,
  "wid": "d3e4f5a6-b7c8-9012-def0-123456789012",
  "tid": "550e8400-e29b-41d4-a716-446655440099",
  "exec_act": "initiate_trade_rollback",
  "par": ["550e8400-e29b-41d4-a716-446655440003"],
  "pol": "compensation_policy_v1",
  "pol_decision": "approved",
  "pol_enforcer": "spiffe://bank.com/human/compliance-officer",
  "compensation_required": true,
  "compensation_reason": "policy_violation_in_parent_trade"
}

The "par" claim links the compensation action to the original trade, creating an auditable chain from execution through violation discovery to remediation.¶

9.4. Autonomous Logistics Coordination

In a logistics workflow, multiple compliance checks must complete before shipment commitment. The DAG structure records that all required checks were completed:¶

Agent A (Route Planning):
  tid: task-001    par: []
  exec_act: plan_route
  pol: route_policy_v1        pol_decision: approved

Agent B (Customs):
  tid: task-002    par: [task-001]
  exec_act: validate_customs
  pol: customs_policy_v2      pol_decision: approved

Agent C (Safety):
  tid: task-003    par: [task-001]
  exec_act: verify_cargo_safety
  pol: safety_policy_v1       pol_decision: approved

Agent D (Payment):
  tid: task-004    par: [task-002, task-003]
  exec_act: authorize_payment
  pol: payment_policy_v3      pol_decision: approved

System (Commitment):
  tid: task-005    par: [task-004]
  exec_act: commit_shipment
  pol: commitment_policy_v1   pol_decision: approved

Note that tasks 002 and 003 both depend only on task-001 and can execute in parallel. Task 004 depends on both, demonstrating the DAG's ability to represent parallel execution with a join point.¶

10.1. Threat Model

The following threat actors are considered:¶

• Malicious agent (insider threat): An agent within the trust domain that intentionally creates false ECT claims.¶

• Compromised agent (external attacker): An agent whose private key has been obtained by an external attacker.¶

• Ledger tamperer: An entity attempting to modify or delete ledger entries after they have been recorded.¶

• Time manipulator: An entity attempting to manipulate timestamps to alter perceived execution ordering.¶

10.2. Self-Assertion Limitation

ECTs are self-asserted by the executing agent. The agent claims what it did, and this claim is signed with its private key. A compromised or malicious agent could create ECTs with false claims (e.g., setting "pol_decision" to "approved" without actually evaluating the policy).¶

ECTs do not independently verify that:¶

• The claimed execution actually occurred as described¶

• The policy evaluation was correctly performed¶

• The input/output hashes correspond to the actual data processed¶

• The agent faithfully performed the stated action¶

The trustworthiness of ECT claims depends on the trustworthiness of the signing agent. To mitigate single-agent false claims, regulated environments SHOULD use the "witnessed_by" mechanism to include independent third-party observers at critical decision points.¶

10.3. Organizational Prerequisites

ECTs operate within a broader trust framework. The guarantees provided by ECTs are only meaningful when the following organizational controls are in place:¶

• Key management governance: Controls over who provisions agent keys and how keys are protected.¶

• Ledger integrity governance: The ledger is maintained by an entity independent of the workflow agents.¶

• Policy lifecycle management: Policy identifiers in ECTs map to actual, validated policy rules.¶

• Agent deployment governance: Agents are deployed and maintained in a manner that preserves their integrity.¶

10.4. Signature Verification

ECTs MUST be signed with the agent's private key using JWS [RFC7515]. The signature algorithm MUST match the algorithm specified in the agent's WIT. Receivers MUST verify the ECT signature against the WIT public key before processing any claims. Receivers MUST verify that the signing key has not been revoked within the trust domain.¶

If signature verification fails, the ECT MUST be rejected entirely and the failure MUST be logged.¶

Implementations MUST use established JWS libraries and MUST NOT implement custom signature verification.¶

10.5. Replay Attack Prevention

ECTs include short expiration times (RECOMMENDED: 5-15 minutes) to limit the window for replay attacks. The "aud" claim restricts replay to unintended recipients: an ECT intended for Agent B will be rejected by Agent C. The "iat" claim enables receivers to reject ECTs that are too old, even if not yet expired.¶

The DAG structure provides additional replay protection: an ECT referencing parent tasks that already have a recorded child task with the same action can be flagged as a potential replay.¶

Implementations SHOULD maintain a cache of recently-seen "jti" values (when present) to detect replayed ECTs within the expiration window.¶

10.6. Man-in-the-Middle Protection

ECTs do not replace transport-layer security. ECTs MUST be transmitted over TLS or mTLS connections. When used with the WIMSE service-to-service protocol [I-D.ietf-wimse-s2s-protocol], transport security is already established. HTTP Message Signatures [RFC9421] provide an alternative channel binding mechanism.¶

The defense-in-depth model provides:¶

• TLS/mTLS (transport layer): Prevents network-level tampering.¶

• WIT/WPT (WIMSE identity layer): Proves agent identity and request authorization.¶

• ECT (execution accountability layer): Records what the agent did and under what policy.¶

10.7. Key Compromise

If an agent's private key is compromised, an attacker can forge ECTs that appear to originate from that agent. To mitigate this risk:¶

• Implementations SHOULD use short-lived keys and rotate them frequently (hours to days, not months).¶

• Private keys SHOULD be stored in Hardware Security Modules (HSMs) or equivalent secure key storage.¶

• Trust domains MUST support rapid key revocation.¶

• Upon suspected compromise, the trust domain MUST revoke the compromised key and issue a new WIT with a fresh key pair.¶

ECTs signed with a compromised key that were recorded in the ledger before revocation remain valid historical records but SHOULD be flagged in the ledger as "signed with subsequently revoked key" for audit purposes.¶

10.8. Collusion and False Claims

A single malicious agent cannot forge parent task references because DAG validation requires parent tasks to exist in the ledger. However, multiple colluding agents could potentially create a false execution history if they control the ledger.¶

Mitigations include:¶

• Independent ledger maintenance: The ledger SHOULD be maintained by an entity independent of the workflow agents.¶

• Witness attestation: Using the "witnessed_by" claim to include independent third-party observers.¶

• Cross-verification: Multiple independent ledger replicas can be compared for consistency.¶

• Out-of-band audit: External auditors periodically verify ledger contents against expected workflow patterns.¶

10.9. Denial of Service

ECT signature verification is computationally inexpensive (approximately 1ms per ECT on modern hardware for ES256). DAG validation complexity is O(V) where V is the number of ancestor nodes reachable from the parent references; for typical shallow DAGs this is efficient.¶

Implementations SHOULD apply rate limiting at the API layer to prevent excessive ECT submissions. DAG validation SHOULD be performed after signature verification to avoid wasting resources on unsigned or incorrectly signed tokens.¶

10.10. Timestamp Accuracy

ECTs rely on timestamps ("iat", "exp") for temporal ordering. Clock skew between agents can lead to incorrect ordering judgments. Implementations SHOULD use synchronized time sources (e.g., NTP) and SHOULD allow a configurable clock skew tolerance (RECOMMENDED: 30 seconds).¶

The temporal ordering check in DAG validation incorporates the clock skew tolerance to account for minor clock differences between agents.¶

10.11. ECT Size Constraints

ECTs with many parent tasks or large extension objects can increase HTTP header size. Implementations SHOULD limit the "par" array to a reasonable size and SHOULD set maximum size limits for the "ext" object to prevent abuse.¶

11.1. Data Exposure in ECTs

ECTs necessarily reveal:¶

• Agent identities ("iss", "aud") for accountability purposes¶

• Action descriptions ("exec_act") for audit trail completeness¶

• Policy evaluation outcomes ("pol", "pol_decision") for compliance verification¶

• Timestamps ("iat", "exp") for temporal ordering¶

ECTs are designed to NOT reveal:¶

• Actual input or output data values (replaced with cryptographic hashes via "inp_hash" and "out_hash")¶

• Internal computation details or intermediate steps¶

• Proprietary algorithms or intellectual property¶

• Personally identifiable information (PII)¶

11.2. Data Minimization

Implementations SHOULD minimize the information included in ECTs. The "exec_act" claim SHOULD use structured identifiers (e.g., "process_payment") rather than natural language descriptions. The "pol" claim SHOULD reference policy identifiers rather than embedding policy content.¶

11.3. Storage and Access Control

ECTs stored in audit ledgers SHOULD be access-controlled so that only authorized auditors and regulators can read them. Implementations SHOULD consider encryption at rest for ledger storage containing sensitive regulatory data.¶

Full input and output data (corresponding to the hashes in ECTs) SHOULD be stored separately from the ledger with additional access controls, since auditors may need to verify hash correctness but general access to the data values is not needed.¶

11.4. Regulatory Access

ECTs are designed for interpretation by qualified human auditors and regulators. ECTs provide structural records of execution ordering and policy evaluation; they are not intended for public disclosure.¶

12.1. Media Type Registration

This document requests registration of the following media type in the "Media Types" registry maintained by IANA:¶

Type name:

application¶

Subtype name:

wimse-exec+jwt¶

Required parameters:

none¶

Optional parameters:

none¶

Encoding considerations:

8bit; an ECT is a JWT that is a JWS using the Compact Serialization, which is a sequence of Base64url-encoded values separated by period characters.¶

Security considerations:

See the Security Considerations section of this document.¶

Interoperability considerations:

none¶

Published specification:

This document¶

Applications that use this media type:

Applications that implement regulated agentic workflows requiring execution context tracing and audit trails.¶

Additional information:

Magic number(s): none File extension(s): none Macintosh file type code(s): none¶

Person and email address to contact for further information:

Christian Nennemann, ietf@nennemann.de¶

Intended usage:

COMMON¶

Restrictions on usage:

none¶

Author:

Christian Nennemann¶

Change controller:

IETF¶

12.2. HTTP Header Field Registration

This document requests registration of the following header field in the "Hypertext Transfer Protocol (HTTP) Field Name Registry" maintained by IANA:¶

Field name:

Execution-Context¶

Status:

permanent¶

Specification document:

This document, Section 5¶

12.3. JWT Claims Registration

This document requests registration of the following claims in the "JSON Web Token Claims" registry maintained by IANA:¶

13.1. Normative References

[I-D.ietf-wimse-arch]

Salowey, J. A., Rosomakho, Y., and H. Tschofenig, "Workload Identity in a Multi System Environment (WIMSE) Architecture", Work in Progress, Internet-Draft, draft-ietf-wimse-arch-06, 30 September 2025, <https://datatracker.ietf.org/doc/html/draft-ietf-wimse-arch-06>.

[I-D.ietf-wimse-s2s-protocol]

Campbell, B., Salowey, J. A., Schwenkschuster, A., and Y. Sheffer, "WIMSE Workload-to-Workload Authentication", Work in Progress, Internet-Draft, draft-ietf-wimse-s2s-protocol-07, 16 October 2025, <https://datatracker.ietf.org/doc/html/draft-ietf-wimse-s2s-protocol-07>.

[RFC2119]

Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, <https://www.rfc-editor.org/rfc/rfc2119>.

[RFC3339]

Klyne, G. and C. Newman, "Date and Time on the Internet: Timestamps", RFC 3339, DOI 10.17487/RFC3339, July 2002, <https://www.rfc-editor.org/rfc/rfc3339>.

[RFC7515]

Jones, M., Bradley, J., and N. Sakimura, "JSON Web Signature (JWS)", RFC 7515, DOI 10.17487/RFC7515, May 2015, <https://www.rfc-editor.org/rfc/rfc7515>.

[RFC7518]

Jones, M., "JSON Web Algorithms (JWA)", RFC 7518, DOI 10.17487/RFC7518, May 2015, <https://www.rfc-editor.org/rfc/rfc7518>.

[RFC7519]

Jones, M., Bradley, J., and N. Sakimura, "JSON Web Token (JWT)", RFC 7519, DOI 10.17487/RFC7519, May 2015, <https://www.rfc-editor.org/rfc/rfc7519>.

[RFC8174]

Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017, <https://www.rfc-editor.org/rfc/rfc8174>.

[RFC9110]

Fielding, R., Ed., Nottingham, M., Ed., and J. Reschke, Ed., "HTTP Semantics", STD 97, RFC 9110, DOI 10.17487/RFC9110, June 2022, <https://www.rfc-editor.org/rfc/rfc9110>.

[RFC9562]

Davis, K., Peabody, B., and P. Leach, "Universally Unique IDentifiers (UUIDs)", RFC 9562, DOI 10.17487/RFC9562, May 2024, <https://www.rfc-editor.org/rfc/rfc9562>.

13.2. Informative References

[DORA]

European Parliament and Council of the European Union, "Regulation (EU) 2022/2554 on digital operational resilience for the financial sector (DORA)", 14 December 2022, <https://eur-lex.europa.eu/eli/reg/2022/2554>.

[EU-AI-ACT]

European Parliament and Council of the European Union, "Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act)", 13 June 2024, <https://eur-lex.europa.eu/eli/reg/2024/1689>.

[EU-MDR]

European Parliament and Council of the European Union, "Regulation (EU) 2017/745 on medical devices (MDR)", 5 April 2017, <https://eur-lex.europa.eu/eli/reg/2017/745>.

[FDA-21CFR11]

U.S. Food and Drug Administration, "Title 21, Code of Federal Regulations, Part 11: Electronic Records; Electronic Signatures", <https://www.ecfr.gov/current/title-21/chapter-I/subchapter-A/part-11>.

[I-D.ietf-scitt-architecture]

Birkholz, H., Delignat-Lavaud, A., Fournet, C., Deshpande, Y., and S. Lasker, "An Architecture for Trustworthy and Transparent Digital Supply Chains", Work in Progress, Internet-Draft, draft-ietf-scitt-architecture-22, 10 October 2025, <https://datatracker.ietf.org/doc/html/draft-ietf-scitt-architecture-22>.

[I-D.ni-wimse-ai-agent-identity]

Yuan, N. and P. C. Liu, "WIMSE Applicability for AI Agents", Work in Progress, Internet-Draft, draft-ni-wimse-ai-agent-identity-01, 20 October 2025, <https://datatracker.ietf.org/doc/html/draft-ni-wimse-ai-agent-identity-01>.

[MIFID-II]

European Parliament and Council of the European Union, "Directive 2014/65/EU of the European Parliament and of the Council on markets in financial instruments (MiFID II)", 15 May 2014, <https://eur-lex.europa.eu/eli/dir/2014/65>.

[OPENTELEMETRY]

Cloud Native Computing Foundation, "OpenTelemetry Specification", <https://opentelemetry.io/docs/specs/otel/>.

[RFC3552]

Rescorla, E. and B. Korver, "Guidelines for Writing RFC Text on Security Considerations", BCP 72, RFC 3552, DOI 10.17487/RFC3552, July 2003, <https://www.rfc-editor.org/rfc/rfc3552>.

[RFC7517]

Jones, M., "JSON Web Key (JWK)", RFC 7517, DOI 10.17487/RFC7517, May 2015, <https://www.rfc-editor.org/rfc/rfc7517>.

[RFC8693]

Jones, M., Nadalin, A., Campbell, B., Ed., Bradley, J., and C. Mortimore, "OAuth 2.0 Token Exchange", RFC 8693, DOI 10.17487/RFC8693, January 2020, <https://www.rfc-editor.org/rfc/rfc8693>.

[RFC9421]

Backman, A., Ed., Richer, J., Ed., and M. Sporny, "HTTP Message Signatures", RFC 9421, DOI 10.17487/RFC9421, February 2024, <https://www.rfc-editor.org/rfc/rfc9421>.

[SPIFFE]

"Secure Production Identity Framework for Everyone (SPIFFE)", <https://spiffe.io/docs/latest/spiffe-about/overview/>.

WIMSE Workload Identity

The WIMSE architecture [I-D.ietf-wimse-arch] and service-to- service protocol [I-D.ietf-wimse-s2s-protocol] provide the identity foundation upon which ECTs are built. WIT/WPT answer "who is this agent?" and "does it control the claimed key?" while ECTs record "what did this agent do?" and "what policy was evaluated?" Together they form an identity-plus-accountability framework for regulated agentic systems.¶

OAuth 2.0 Token Exchange

[RFC8693] defines the OAuth 2.0 Token Exchange protocol and registers the "act" (Actor) claim in the JWT Claims registry. ECTs intentionally use the distinct claim name "exec_act" for the action/task type to avoid collision with the "act" claim. Transaction tokens in OAuth establish API authorization context; ECTs serve the complementary purpose of recording execution accountability across multi-step workflows.¶

Distributed Tracing (OpenTelemetry)

OpenTelemetry [OPENTELEMETRY] and similar distributed tracing systems provide observability for debugging and monitoring. ECTs differ in several important ways: ECTs are cryptographically signed per-task with the agent's private key; ECTs are tamper-evident through JWS signatures; ECTs enforce DAG validation rules; and ECTs are designed for regulatory audit rather than operational monitoring. OpenTelemetry data is typically controlled by the platform operator and can be modified or deleted without detection. ECTs and distributed traces are complementary: traces provide observability while ECTs provide signed execution records. ECTs may reference OpenTelemetry trace identifiers in the "ext" claim for correlation.¶

Blockchain and Distributed Ledgers

Both ECTs and blockchain systems provide immutable records. This specification intentionally defines only the ECT token format and is agnostic to the storage mechanism. ECTs can be stored in append-only logs, databases with cryptographic commitments, blockchain networks, or any storage providing the required properties defined in Section 8.¶

SCITT (Supply Chain Integrity, Transparency, and Trust)

The SCITT architecture [I-D.ietf-scitt-architecture] defines a framework for creating transparent and auditable supply chain records through Transparency Services, Signed Statements, and Receipts. ECTs and SCITT are naturally complementary: the ECT "wid" (Workflow Identifier) claim can serve as a correlation identifier referenced in SCITT Signed Statements, linking a complete ECT audit trail to a supply chain transparency record. For example, in a regulated manufacturing workflow, each agent step produces an ECT (recording what was done, by whom, under what policy), while the overall workflow identified by "wid" is registered as a SCITT Signed Statement on a Transparency Service. This enables auditors to verify both the individual execution steps (via ECT DAG validation) and the end-to-end supply chain integrity (via SCITT Receipts) using the "wid" as the shared correlation point. The "ext" claim in ECTs (Section 4.2.2) can carry SCITT-specific metadata such as Transparency Service identifiers or Receipt references for tighter integration.¶

W3C Verifiable Credentials

W3C Verifiable Credentials represent claims about subjects (e.g., identity, qualifications). ECTs represent execution records of actions (what happened, in what order, under what policy). While both use JWT/JWS as a serialization format, their semantics and use cases are distinct.¶

Minimal Implementation

A minimal conforming implementation should:¶

1. Create JWTs with all required claims ("iss", "aud", "iat", "exp", "tid", "exec_act", "par", "pol", "pol_decision").¶

2. Sign ECTs with the agent's private key using an algorithm matching the WIT (ES256 recommended).¶

3. Verify ECT signatures against WIT public keys.¶

4. Perform DAG validation (parent existence, temporal ordering, cycle detection).¶

5. Append verified ECTs to an audit ledger.¶

Storage Recommendations

• Append-only log: Simplest approach; immutability by design.¶

• Database with hash chains: Periodic cryptographic commitments over batches of entries.¶

• Distributed ledger: Maximum immutability guarantees for cross-organizational audit.¶

• Hybrid: Hot storage in a database, cold archive in immutable storage.¶

Performance Considerations

• ES256 signature verification: approximately 1ms per ECT on modern hardware.¶

• DAG validation: O(V) where V is the number of reachable ancestor nodes (typically small for shallow workflows).¶

• JSON serialization: sub-millisecond per ECT.¶

• Total per-request overhead: approximately 5-10ms, acceptable for regulated workflows where correctness is prioritized over latency.¶

Interoperability

Implementations should use established JWT/JWS libraries (JOSE) for token creation and verification. Custom cryptographic implementations should not be used. Implementations should be tested against multiple JWT libraries to ensure interoperability.¶

Example 1: Simple Two-Agent Workflow

Agent A executes a data retrieval task and sends the ECT to Agent B:¶

ECT JOSE Header:¶

{
  "alg": "ES256",
  "typ": "wimse-exec+jwt",
  "kid": "agent-a-key-2026-02"
}

ECT Payload:¶

{
  "iss": "spiffe://example.com/agent/data-retrieval",
  "sub": "spiffe://example.com/agent/data-retrieval",
  "aud": "spiffe://example.com/agent/validator",
  "iat": 1772064150,
  "exp": 1772064750,
  "wid": "b1c2d3e4-f5a6-7890-bcde-f01234567890",
  "tid": "550e8400-e29b-41d4-a716-446655440001",
  "exec_act": "fetch_patient_data",
  "par": [],
  "pol": "clinical_data_access_policy_v1",
  "pol_decision": "approved",
  "inp_hash": "sha-256:n4bQgYhMfWWaL-qgxVrQFaO_TxsrC4Is0V1sFbDwCgg",
  "out_hash": "sha-256:LCa0a2j_xo_5m0U8HTBBNBNCLXBkg7-g-YpeiGJm564",
  "exec_time_ms": 142,
  "regulated_domain": "medtech"
}

Agent B receives the ECT, verifies it, executes a validation task, and creates its own ECT:¶

{
  "iss": "spiffe://example.com/agent/validator",
  "sub": "spiffe://example.com/agent/validator",
  "aud": "spiffe://example.com/system/ledger",
  "iat": 1772064160,
  "exp": 1772064760,
  "wid": "b1c2d3e4-f5a6-7890-bcde-f01234567890",
  "tid": "550e8400-e29b-41d4-a716-446655440002",
  "exec_act": "validate_safety",
  "par": ["550e8400-e29b-41d4-a716-446655440001"],
  "pol": "safety_validation_policy_v2",
  "pol_decision": "approved",
  "exec_time_ms": 89,
  "regulated_domain": "medtech"
}

The resulting DAG:¶

task-...-0001 (fetch_patient_data)
  |
  v
task-...-0002 (validate_safety)

Example 2: Medical Device SDLC with Release Approval

A multi-step medical device software lifecycle workflow with autonomous agents and human release approval:¶

Task 1 (Spec Review Agent):¶

{
  "iss": "spiffe://meddev.example/agent/spec-reviewer",
  "sub": "spiffe://meddev.example/agent/spec-reviewer",
  "aud": "spiffe://meddev.example/agent/code-gen",
  "iat": 1772064150,
  "exp": 1772064750,
  "wid": "c2d3e4f5-a6b7-8901-cdef-012345678901",
  "tid": "a1b2c3d4-0001-0000-0000-000000000001",
  "exec_act": "review_requirements_spec",
  "par": [],
  "pol": "spec_review_policy_v2",
  "pol_decision": "approved",
  "regulated_domain": "medtech",
  "model_version": "spec-review-v3.1",
  "inp_hash": "sha-256:n4bQgYhMfWWaL-qgxVrQFaO_TxsrC4Is0V1sFbDwCgg",
  "out_hash": "sha-256:LCa0a2j_xo_5m0U8HTBBNBNCLXBkg7-g-YpeiGJm564"
}

Task 2 (Code Generation Agent):¶

{
  "iss": "spiffe://meddev.example/agent/code-gen",
  "sub": "spiffe://meddev.example/agent/code-gen",
  "aud": "spiffe://meddev.example/agent/test-runner",
  "iat": 1772064200,
  "exp": 1772064800,
  "wid": "c2d3e4f5-a6b7-8901-cdef-012345678901",
  "tid": "a1b2c3d4-0001-0000-0000-000000000002",
  "exec_act": "implement_module",
  "par": ["a1b2c3d4-0001-0000-0000-000000000001"],
  "pol": "coding_standards_v3",
  "pol_decision": "approved",
  "regulated_domain": "medtech",
  "model_version": "codegen-v2.4"
}

Task 3 (Autonomous Test Agent):¶

{
  "iss": "spiffe://meddev.example/agent/test-runner",
  "sub": "spiffe://meddev.example/agent/test-runner",
  "aud": "spiffe://meddev.example/agent/build",
  "iat": 1772064260,
  "exp": 1772064860,
  "wid": "c2d3e4f5-a6b7-8901-cdef-012345678901",
  "tid": "a1b2c3d4-0001-0000-0000-000000000003",
  "exec_act": "execute_test_suite",
  "par": ["a1b2c3d4-0001-0000-0000-000000000002"],
  "pol": "test_coverage_policy_v1",
  "pol_decision": "approved",
  "regulated_domain": "medtech",
  "exec_time_ms": 4523
}

Task 4 (Build Agent):¶

{
  "iss": "spiffe://meddev.example/agent/build",
  "sub": "spiffe://meddev.example/agent/build",
  "aud": "spiffe://meddev.example/human/release-mgr-42",
  "iat": 1772064310,
  "exp": 1772064910,
  "wid": "c2d3e4f5-a6b7-8901-cdef-012345678901",
  "tid": "a1b2c3d4-0001-0000-0000-000000000004",
  "exec_act": "build_release_artifact",
  "par": ["a1b2c3d4-0001-0000-0000-000000000003"],
  "pol": "build_validation_v2",
  "pol_decision": "approved",
  "regulated_domain": "medtech",
  "out_hash": "sha-256:Ry1YfOoW2XpC5Mq8HkGzNx3dL9vBa4sUjE7iKt0wPZc"
}

Task 5 (Human Release Manager Approval):¶

{
  "iss": "spiffe://meddev.example/human/release-mgr-42",
  "sub": "spiffe://meddev.example/human/release-mgr-42",
  "aud": "spiffe://meddev.example/system/ledger",
  "iat": 1772064510,
  "exp": 1772065110,
  "wid": "c2d3e4f5-a6b7-8901-cdef-012345678901",
  "tid": "a1b2c3d4-0001-0000-0000-000000000005",
  "exec_act": "approve_release",
  "par": ["a1b2c3d4-0001-0000-0000-000000000004"],
  "pol": "release_approval_policy",
  "pol_decision": "approved",
  "pol_enforcer": "spiffe://meddev.example/human/release-mgr-42",
  "witnessed_by": [
    "spiffe://meddev.example/audit/qa-observer-1"
  ],
  "regulated_domain": "medtech"
}

The resulting DAG records the complete SDLC: spec review preceded implementation, implementation preceded testing, testing preceded build, and a human release manager approved the final release with independent witness attestation.¶

task-...-0001 (review_requirements_spec)
  |
  v
task-...-0002 (implement_module)
  |
  v
task-...-0003 (execute_test_suite)
  |
  v
task-...-0004 (build_release_artifact)
  |
  v
task-...-0005 (approve_release) [human, witnessed]

An FDA auditor reconstructs this DAG by querying the audit ledger for all ECTs with wid "c2d3e4f5-a6b7-8901-cdef-012345678901" and verifying each signature. The DAG provides cryptographic evidence that the SDLC followed the prescribed process with human oversight at the release gate.¶

Example 3: Parallel Execution with Join

A workflow where two tasks execute in parallel and a third task depends on both:¶

task-...-0001 (assess_risk)
  |            \
  v             v
task-...-0002  task-...-0003
(check         (verify
compliance)    liquidity)
  |            /
  v           v
task-...-0004 (execute_trade)

Task 004 ECT payload:¶

{
  "iss": "spiffe://bank.example/agent/execution",
  "sub": "spiffe://bank.example/agent/execution",
  "aud": "spiffe://bank.example/system/ledger",
  "iat": 1772064250,
  "exp": 1772064850,
  "wid": "d3e4f5a6-b7c8-9012-def0-123456789012",
  "tid": "f1e2d3c4-0004-0000-0000-000000000004",
  "exec_act": "execute_trade",
  "par": [
    "f1e2d3c4-0002-0000-0000-000000000002",
    "f1e2d3c4-0003-0000-0000-000000000003"
  ],
  "pol": "trade_execution_policy_v3",
  "pol_decision": "approved",
  "regulated_domain": "finance"
}

The "par" array with two entries records that both compliance checking and liquidity verification were completed before trade execution.¶