feat(ssh): wave 0 — kernel contract seam (computerId)

Add additive optional `computerId` field to ToolExecuteContext + RunTurnInput.
The kernel never interprets it (forwards verbatim to tools, like cwd) — it never
enters the model prompt (no prompt-cache impact). When omitted/undefined,
execution is LOCAL (today's behavior), so this is fully backward compatible.

This is the orchestrator-authored seam (ORCHESTRATOR.md §2a) that lets Wave 1's
producers (wire Computer types, exec-backend contract) and the consumer
(kernel runtime threading) run in parallel against a fixed type.

Refs: notes/ssh-support-plan.md (decisions resolved in §0.5/§13).
No merge or push.

1 files changed, 12 insertions, 0 deletions

diff --git a/packages/kernel/src/contracts/tool.ts b/packages/kernel/src/contracts/tool.ts
index d5a835c..589fbd0 100644
--- a/packages/kernel/src/contracts/tool.ts
+++ b/packages/kernel/src/contracts/tool.ts
@@ -78,6 +78,18 @@ export interface ToolExecuteContext {
 	 * a tool has no conversation scope (e.g. a global tool).
 	 */
 	readonly conversationId?: string;
+
+	/**
+	 * The computer this tool-call executes on (SSH support). When
+	 * omitted/undefined, execution is LOCAL (today's behavior — the tool uses
+	 * the local node fs/child_process). When set, it is an SSH config alias
+	 * (see `notes/ssh-support-plan.md` §3); a tool resolves a remote
+	 * `ExecBackend` for it via its injected resolver. The kernel never
+	 * interprets it — it forwards the value verbatim from
+	 * `RunTurnInput.computerId`, exactly like `cwd`. It never enters the model
+	 * prompt, so it does not affect prompt caching.
+	 */
+	readonly computerId?: string;
 }
 
 /**