← All agent-stack verdicts

Agent-to-agent protocols (A2A class)

Adopt and wrap Confidence Low Group Orchestration Evidence as of 2026-09-23

Written by VerdictsOrchestration · Reviewed by control-plane pane 2 on 2026-09-23 (separate AI agent sessions; review record)

Bracketed tags are evidence tiers from RULEBOOK.md §1: [Verified] means checked directly, [CI-observed] read from a project’s CI pages, [Maintainer claim] asserted by the project, [External] a dated third-party source, [Inference] our judgment. A gold path:line link opens the cited line. TRL is technology readiness level, scored 1 to 9. CI is continuous integration: automated checks that run on every push. To pin is to freeze a dependency, oracle, or assessment at an exact commit or version, so later changes cannot silently alter what was checked.

Bottom line

Inference, low confidence: if your agents must discover and delegate to agents from other teams, target A2A v1.x through one of its official SDKs, and do not design your own protocol. Wrap it: pin the shared compatibility test kit (the TCK) and run it against your build, test every transport the spec requires, and call any interop result "self-interop" until you have tested against an implementation from outside the A2A organization, because the evidence found none. Confidence is low because the evidence pack itself calls the category thin.

Adopt, do not rebuild

License checked on 2026-09-23; non-permissive licenses are named in the text.

Copy these practices

Build only if

  • No evidenced constraint justifies a new protocol: the category consolidated from two specs to one, and the companion's own charter rules out new protocol design. Building a clean-room implementation of A2A is defensible only where no official SDK exists for your language, and even then the TCK is the bar. Whatever you build or adopt, count a result as full interop only against a peer implementation from outside the originating organization and call anything else self-interop; that is the companion's rule, and the interop tested in this category is inside the A2A organization's own SDKs. [Inference] (ecosystem/pickup/_evidence/multi-agent-protocols.md:33 A new clean-room project should treat A2A v1.x as the target and ACP as historical; ecosystem/pickup/pickup-multi-agent-protocols.md:21 New protocol design (the charter is conformance, not invention); ecosystem/pickup/pickup-multi-agent-protocols.md:44 full INTEROP requires ≥1 independent peer implementation; ecosystem/pickup/_evidence/multi-agent-protocols.md:32 There is no second fully independent A2A implementation outside the a2aproject org found in this sweep)

Where FrankenSuite touches this

  • None of the 44 packets implements an agent-to-agent protocol. The nearest is frankenterm's Robot Mode, a typed JSON envelope with golden test matrices for one agent driving others through terminal panes, which its packet proposes testing against a non-frankenterm backend before calling it a standard. [Verified] (packets/frankenterm-assessment.md:37 is the closest thing in the wild to an agent-to-terminal control contract)
  • franken_alignment models commit-reveal voting rounds among helper agents as an in-memory reference model only; its packet grades it TRL 3 and nothing production exists. [Verified] (packets/franken_alignment-assessment.md:19 salted commit–reveal helper congresses; packets/franken_alignment-assessment.md:196 Technology readiness | TRL 3)
  • The suite's own multi-agent development coordinates through MCP Agent Mail, a development tool the packets describe but did not assess as a product. [Maintainer claim] (packets/franken_nlp-assessment.md:67 MCP Agent Mail coordination, bead-assignee locking)

What we cannot say

Revisit when

  • An A2A implementation from outside the a2aproject organization appears and passes the pinned TCK: interop claims can then move past self-interop, and confidence can rise.
  • All four official SDKs are observed running the TCK in CI.
  • A2A publishes a v2 or a breaking v1.x revision: re-pin the TCK and the compat suites before any claim.

Sources

Starting a build here?

Begin from the pickup companion for Agent-to-agent protocols (A2A class): it carries the charter seed, the oracle candidates and how to verify them, and the first claims to register, so a new project starts from evidence instead of from scratch.

Help fix it

Each opens a GitHub issue form with this page’s address filled in. You need a GitHub account.

Made by Joshua Nowak (ZestStream) with AI coding agents. Not affiliated with or funded by Jeffrey Emanuel; Joshua uses his public tools and is a paying subscriber to jeffreys-skills.md.

Source on GitHub MIT licensed