OpenA2A's specifications are authored in the open, and all but ai-safety.txt and the OTel semantic conventions are published with working reference implementations. They are early, and we are looking for co-authors and contributors to help shape them before they go to external standards bodies. Here is what we are looking for and how to help.
Key idea
Build an implementation
Every specification except ai-safety.txt and the OTel semantic conventions ships with a reference implementation, but interoperability needs independent ones. We are looking for implementations in other languages and stacks: verifiers, clients, SDKs, and brokers. One concrete starting point is open now: making the AAP broker's RFC 8693 token exchange reachable from a shipped daemon (agent-authorization-protocol issue #1). Building a second implementation is one of the most valuable ways to harden a spec.
Security and cryptography researchers, including academic and PhD-level work.
Standards-process experts (W3C, IETF, OpenTelemetry) who can help take these specifications to external bodies.
Engineers building agent platforms and runtimes, for independent implementations and adoption.
Red teamers and security auditors.
What each spec needs help with
One concrete next step per specification. Where an open issue exists, it is linked.
did:opena2a W3C DID experts to review the 0.3 method specification, registered in the W3C DID Extensions registry (w3c/did-extensions#717, merged 2026-07-04).
AIP independent implementations and interoperability testing against OpenID Connect, WebAuthn, and W3C Verifiable Credentials.
ATX a verifier written outside OpenA2A (the reference verifiers are Go and Python in atx-conformance and TypeScript in @opena2a/atx-verify), and cryptography review of the Ed25519 and hybrid ML-DSA-65 signing.
ATP review of the RFC 6962 Merkle transparency-log design and independent log monitors.
AAP making the broker's RFC 8693 token exchange, a library in Secretless today, reachable from a shipped daemon, and delegation-model review. open issue: agent-authorization-protocol#1
ATM validated threat techniques (Observed, Validated, or Adapted evidence).
AIIS high-precision attack signatures and applied-cryptography review.
ABGS governance and compliance experts to review the control set.
OASB security-product teams to run detectors against the benchmark, contribute attack scenarios, and be listed as adopters.
Key idea
These specifications are early by design
We are inviting co-authors to shape them now, before they go to external standards bodies. Reviews, second implementations, and security audits at this stage are worth far more than they will be after a spec is frozen. If you are not sure where to start, read the five-minute orientation and pick the layer closest to your expertise.