Agent identity on the wire + a true per-agent stop (follow-up)

[hellno]

Part of #179. Follow-up — deferred until after the demo wave. Companion to #203 (Lockdown rename); pairs with the #198 wire work; sequenced after #31.

Problem

ProposalOrigin::Agent is anonymous — the E2 handle is app-local cosmetics, and Policy.revoked is one global flag. So the only honest "stop" today is the global Lockdown (#203). To stop ONE agent — or to show two agents without lying — the wire needs identity.

Scope (spec level)

• ProposalOrigin::Agent { id } (additive variant or field, Wire-contract evolution: Hello capability discovery + the five evolution rules #31 evolution rules — same mechanism as Dapp origin on the wire: attribute browser transactions to the requesting site #198's Dapp variant).
• The MCP sidecar passes a stable per-session agent id; the daemon tracks revocation per agent id.
• New scoped revoke on the wire; RevokeAll (Lockdown) unchanged.
• E6's brake becomes genuinely agent-scoped ("Stop this agent"); Lockdown remains the global action.
• Policy question to settle in-spec: per-agent rules, or shared rules + per-agent revocation only? Start with the latter — smaller, and it answers the demo need (name the agent, stop the agent).

Non-goals

• Multi-agent policy authoring UI.
• Agent authentication/attestation (identity here is a label + revocation key, not a security boundary — say so honestly in the UI).

Acceptance (when picked up)

Two concurrently connected agents render distinctly in Activity; stopping one leaves the other running; Lockdown still stops both. Wire round-trip + evolution tests.