Abstract

Agent-to-agent delegation is now production infrastructure. Agent protocols model work as long-running tasks with streamed updates, artifacts, push notifications and cancellation. Yet systems often confuse four pairs of things:

  • a proposal with an accepted agreement;
  • a receipt with acceptance;
  • a terminal stream marker with successful work;
  • a cancellation with reversal of earlier effects.

This paper defines evidence-qualified consistency reconstruction for incrementally negotiated and streamed tasks. It keeps three distinctions:

  • task proposals versus accepted agreement versions;
  • task status versus stream closure, artifact finalisation and observable external effects;
  • subactions already executed versus obligations that remain after amendments or cancellation.

A six-phase, 44-stage workflow returns CONSISTENT, INCONSISTENT or UNRESOLVED through a scoped three-valued conjunction, in which a contradiction dominates and missing evidence is never treated as a negative result.

The paper presents:

  • eleven inventive steps, covering causal event ordering, field-scoped partial acceptance, stale-base and concurrent-revision branch preservation, version-bound results, streamed-artifact finality, cancellation-race interval comparison, partition and replay reconciliation, dependency-aware residual work, and scoped closure;
  • six embodiments;
  • forty implementation checkpoints;
  • a synthetic laboratory in Python, Go and JavaScript (52 fixtures, 166 passing test executions, about 0.16 to 2.0 microseconds per in-memory evaluation). The laboratory is self-authored and not independently or field validated.

An agreement assessment grants no authority to execute a consequential action.

A companion industry-relevance section aligns the method with public roadmaps:

  • Google: the Agent2Agent (A2A) protocol (task states, Server-Sent Events streaming, push notifications), Agent Development Kit and native A2A support in its agent platform.
  • Microsoft: Agent 365 with cross-cloud registry sync, A2A in Foundry Agent Service, Agent Framework and Copilot Studio.
  • Others: A2A founding members AWS, Cisco, Salesforce, SAP and ServiceNow, and Model Context Protocol (MCP) based agent tooling.

Disclaimer: The industry and product-roadmap relevance is an AI-generated alignment based only on public announcements. It is not an endorsement, partnership, sponsorship, licence, technical validation or evaluation by Google, Microsoft, the Linux Foundation or any other company or standards body. It does not state or imply that any company uses, needs, has assessed or infringes this method or any patent application. All company, product and protocol names are trademarks of their respective owners and are used only to describe technical relevance. Patent pending: Indian Patent Application No. 202631129178 (source disclosure on Zenodo, record 23793543). Related PCT application: PCT/IB2026/055615, filed 4 June 2026 and publicly disclosed 16 July 2026. Readers must read the original disclosures before drawing any conclusion. The inventor is open to strategic collaboration or acquihire discussions.

Keywords

AI safety, multi-agent systems, agent-to-agent protocol, A2A, MCP, task negotiation, agreement versioning, partial acceptance, stale-base revision, concurrent amendments, streaming results, Server-Sent Events, push notifications, artifact finalisation, cancellation race, late results, partition recovery, replay, residual obligations, causal ordering, three-valued logic, closure verdict, Google A2A, Agent Development Kit, Microsoft Agent 365, Foundry Agent Service, Agent Framework

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License

Download

Share

COinS