Most agent identity work assumes agents already share infrastructure: a company directory, a shared registry, a delegation chain within one system. A November 2025 paper out of TU Berlin's Service-centric Networking group starts from a narrower, sharper problem: what happens when two agents meet for the first time, with no prior relationship, possibly across organizational boundaries? The paper's framing is blunt about the gap. LLM-based agents currently have no way to build differentiated trust at the start of a dialogue with an unfamiliar peer — every counterpart looks the same until something goes wrong. The authors argue this becomes a real liability once agents stop operating in isolated, single-vendor environments and start negotiating, delegating tasks, or exchanging data across companies. Their proposed fix combines two existing W3C-track technologies rather than inventing new cryptography: long-lived Decentralized Identifiers (DIDs) as the agent's persistent handle, and Verifiable Credentials (VCs) as tamper-proof, third-party-issued attestations bound to that identity — things like capability scope, provenance, or behavioral history. The idea is that an agent shows up to a new interaction carrying credentials another party can check without needing to phone home to a central directory or trust the agent's own claims about itself. The paper builds this into a working prototype: a multi-agent system where each participant is given a self-sovereign digital identity rather than a token issued by whichever platform happens to host it. What makes this worth flagging separately from the broader DID/VC literature already circulating (LOKA's ethics-layer pitch, Huang et al.'s zero-trust IAM framework, CSA's blueprint) is the specific target: not governance-by-design or directory-scale discovery, but the narrow mechanics of a cold-start handshake between agents with no shared history. The paper positions this as filling a gap left by related work — schemes like authenticated delegation among agents that rely on VCs or OIDC still presuppose some existing relationship or session context to bootstrap from. Self-sovereign identity, bound to portable attestations, is offered as the piece that lets trust travel with the agent instead of staying pinned to whoever it's currently talking to. This is early-stage academic work, not a shipping standard, and the authors frame it as a conceptual framework plus prototype rather than a production system. But it's a useful marker of where the identity conversation is heading next: past 'does this agent belong to this company' and toward 'can a total stranger verify anything about this agent in the first ten seconds of contact.'