run: fold inv.Images into the initial user message (multimodal opening turn)

The executor passed only the text `input` to majordomo's agent.Run, silently
dropping inv.Images — so a multimodal run (vision: chatbot @mention, chat API)
lost its images on the executus path. majordomo's Run input arg is text-only, so
fold the images into the first user message (text + image parts) via WithHistory
and call Run with empty input, mirroring mort agentexec's multimodal seeding. The
image-less path is unchanged (prompt passes straight through).

Tests: a run with Images carries the image bytes + prompt into the first model
request; the text-only path still reaches the model.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

run/executor.go

@@ -320,7 +320,7 @@ func (e *Executor) Run(ctx context.Context, ra RunnableAgent, inv tool.Invocatio
 	// One WithSteer drains BOTH the session mailbox (a tool's AttachImages) and
 	// the critic's nudges before each step.
 	steer := func() []llm.Message { return append(mailbox.drain(), critic.drainSteer()...) }
-	runRes, runErr := ag.Run(runCtx, input, agent.WithSteer(steer))
+	runRes, runErr := runAgent(runCtx, ag, input, inv.Images, agent.WithSteer(steer))
 
 	status := statusFor(runCtx, runErr)
 	if runRes != nil {
@@ -440,3 +440,21 @@ func detach(ctx context.Context) context.Context {
 	_ = cancel // bounded by the timeout; nothing to cancel early
 	return c
 }
+
+// runAgent dispatches the majordomo agent loop. majordomo's Run takes a text-only
+// input arg, so when the invocation carries images they're folded into the first
+// user message (text + image parts) via WithHistory and Run is called with an
+// empty input — the model then sees a multimodal opening turn. The image-less path
+// passes the prompt straight through. Mirrors mort agentexec's multimodal seeding.
+func runAgent(ctx context.Context, ag *agent.Agent, input string, images []llm.ImagePart, opts ...agent.RunOption) (*agent.Result, error) {
+	if len(images) == 0 {
+		return ag.Run(ctx, input, opts...)
+	}
+	parts := make([]llm.Part, 0, len(images)+1)
+	parts = append(parts, llm.Text(input))
+	for _, img := range images {
+		parts = append(parts, img)
+	}
+	opts = append(opts, agent.WithHistory([]llm.Message{llm.UserParts(parts...)}))
+	return ag.Run(ctx, "", opts...)
+}