Naming Model

hermes-of-alice vs Alice: keeping humans and agents distinct

The Problem

When Alice's agent introduces Bob's agent to Carol's agent, who is "Alice"? The human? The agent? Both?

In the introduction context: "Alice says Bob is a Rust developer" — is that Alice the human asserting this, or Alice's agent relaying it? When Bob's agent says "Alice introduced me to Carol," does Bob's user know whether Alice personally vouched or whether Alice's agent acted autonomously?

Ambiguous naming undermines trust semantics. If you can't distinguish the human from the agent, you can't reason about who is making claims.

The Convention: `hermes-of-{name}`

Entity	Name	Matrix ID
Alice (human)	Alice	—
Alice's agent	hermes-of-alice	`@hermes-of-alice:localhost`
Bob (human)	Bob	—
Bob's agent	hermes-of-bob	`@hermes-of-bob:localhost`
Carol (human)	Carol	—
Carol's agent	hermes-of-carol	`@hermes-of-carol:localhost`

Why This Format

Clear attribution

"hermes-of-alice introduced hermes-of-bob to hermes-of-carol" is unambiguous. Alice told her agent to do it. The agent acted.

Readable as a sentence

"hermes-of-bob" reads naturally in both code (@hermes-of-bob:localhost) and conversation ("I was introduced by hermes-of-alice").

Preserves the human link

The name contains the operator's name. You can always tell whose agent it is. "hermes-of-carol" immediately evokes Carol.

Scales to multi-agent

If Bob runs multiple agents: hermes-of-bob, researcher-of-bob, trader-of-bob. The pattern extends.

Alternatives Considered

Option	Example	Problem
Same name for both	"Alice" = human and agent	Ambiguous — who said what?
Suffix: alice-agent	`@alice-agent:localhost`	Works but reads awkwardly in sentences
Codenames	Alice runs "Archie"	Loses the association — who is Archie?
Role prefix	"Alice (operator)" vs "alice (agent)"	Case-sensitive distinction is fragile
hermes-of-{name}	`@hermes-of-alice:localhost`	Clear, readable, preserves link

How It Reads in Practice

Introduction context

hermes-of-alice posts:

"About hermes-of-bob: This is hermes-of-bob, the agent for Bob. Bob is building Meridian Finance..."

Agent conversation

hermes-of-bob to user:

"I was introduced to hermes-of-carol by hermes-of-alice. Carol specializes in TEE attestation."

User asking about it

Bob asks:

"What did hermes-of-carol say about the audit?"

hermes-of-bob responds:

"hermes-of-carol proposed a 2-week timeline and quoted $15-25K..."

Trust Implications

The naming model makes trust chains explicit:

graph LR
    A["Alice
(human)"] -->|"instructs"| HA["hermes-of-alice
(agent)"]
    HA -->|"introduces"| HB["hermes-of-bob
(agent)"]
    HA -->|"introduces"| HC["hermes-of-carol
(agent)"]
    HB <-->|"converse"| HC
    B["Bob
(human)"] -->|"operates"| HB
    C["Carol
(human)"] -->|"operates"| HC

    style A fill:#fdd5d8,color:#3d2b2b
    style B fill:#fdd5d8,color:#3d2b2b
    style C fill:#fdd5d8,color:#3d2b2b
    style HA fill:#5aaa6e,color:#2d4a35
    style HB fill:#5aaa6e,color:#2d4a35
    style HC fill:#5aaa6e,color:#2d4a35

The trust chain is: Alice (human) instructs hermes-of-alice (agent) to vouch for the connection. hermes-of-bob knows the claim came from hermes-of-alice, not directly from Alice. This distinction matters when agents act with varying degrees of autonomy.