docs(comments): 注释中文化 · internal/runtime/orchestration（29 文件，纯替换）

- 只翻注释/文档字符串：剥离注释后代码行逐字节不变（机械校验只剩允许的行尾注释形态）
- 裁决编号与 design 节号（02.4.3 / 11.4.2 / design/05 §9 等）逐字保留
- 复核：go build 绿；orchestration 及其子包（audit/comm）测试绿；gofmt 干净

internal/runtime/orchestration/spawntool_test.go

@@ -18,8 +18,8 @@ import (
 	"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/tool"
 )
 
-// countingRunner is a scheduler runner that records how many invocations overlap,
-// so a test can prove the batch still goes through the scheduler's parallel gate.
+// countingRunner 是一个调度器 runner，记录有多少次调用互相重叠，使测试能证明该批次
+// 仍走调度器的并行闸门。
 type countingRunner struct {
 	delay  time.Duration
 	result *loop.Result
@@ -64,16 +64,15 @@ func (r *countingRunner) peak() int {
 	return r.maxActive
 }
 
-// spawnToolFor builds a dispatch tool over a scheduler with the given options and
-// foreground window.
+// spawnToolFor 在带给定选项与前台窗口的调度器上构建派发工具。
 func spawnToolFor(t *testing.T, opts Options, wait time.Duration) (*Scheduler, *SpawnTool) {
 	t.Helper()
 	s := newScheduler(t, opts)
 	return s, NewSpawnTool(s, WithSpawnWait(wait))
 }
 
-// decodeSpawnResult re-reads the tool result through JSON, so a test asserts the
-// shape the model actually receives rather than the in-memory struct.
+// decodeSpawnResult 经 JSON 重新读取工具结果，使测试断言的是模型实际收到的形状，而不是
+// 内存里的结构体。
 func decodeSpawnResult(t *testing.T, raw any) SpawnResult {
 	t.Helper()
 	b, err := json.Marshal(raw)
@@ -87,7 +86,7 @@ func decodeSpawnResult(t *testing.T, raw any) SpawnResult {
 	return out
 }
 
-// statusOf returns the status the scheduler reports for id, or "" when unknown.
+// statusOf 返回调度器为 id 报告的状态；未知时返回 ""。
 func statusOf(s *Scheduler, id string) Status {
 	for _, a := range s.Agents() {
 		if a.ID == id {
@@ -159,8 +158,7 @@ func TestSpawnToolParallelBatchHonorsParallelLimit(t *testing.T) {
 			t.Fatalf("agent %s status = %q, want done", a.ID, a.Status)
 		}
 	}
-	// The batch went through the scheduler's semaphore: with MaxParallel=1 the
-	// three sub-agents can never run at the same time.
+	// 该批次确实走过了调度器的信号量：MaxParallel=1 时三个子代理绝不可能同时运行。
 	if peak := r.peak(); peak != 1 {
 		t.Fatalf("max concurrent runs = %d, want 1 (the parallel limit)", peak)
 	}
@@ -247,7 +245,7 @@ func TestSpawnToolMovesSlowBatchToBackground(t *testing.T) {
 func TestSpawnToolKeepsPerAgentTimeout(t *testing.T) {
 	t.Parallel()
 	s := newScheduler(t, Options{MaxParallel: 1, PerAgentTimeout: 40 * time.Millisecond})
-	s.runner = &mockRunner{gate: make(chan struct{})} // never closed: block until the timeout
+	s.runner = &mockRunner{gate: make(chan struct{})} // 永不关闭：阻塞到超时
 	sp := NewSpawnTool(s, WithSpawnWait(2*time.Second))
 
 	raw, err := sp.Execute(context.Background(), map[string]any{"goal": "hang"})
@@ -337,9 +335,8 @@ func TestSpawnToolWithoutScheduler(t *testing.T) {
 	}
 }
 
-// TestSpawnToolThroughDispatcher drives the tool through the real dispatcher so the
-// advertised JSON Schema, the argument validation and the registry's reachability
-// markers are all exercised — not just Execute.
+// TestSpawnToolThroughDispatcher 让该工具走真实的 dispatcher，使对外广告的 JSON
+// Schema、参数校验与注册表的可达性标记都被演练到 —— 而不只是 Execute。
 func TestSpawnToolThroughDispatcher(t *testing.T) {
 	t.Parallel()
 	s, sp := spawnToolFor(t, Options{MaxParallel: 2}, time.Second)
@@ -374,8 +371,8 @@ func TestSpawnToolThroughDispatcher(t *testing.T) {
 		t.Fatalf("result = %+v, want one settled agent", got)
 	}
 
-	// Registered with the default (both false) markers: owner-only, so a group or
-	// stranger session can neither see nor invoke it (design/11 §3).
+	// 以默认标记（两个都 false）注册：仅主人可用，因此群聊或陌生人会话既看不到也
+	// 调不动它（design/11 §3）。
 	if !restrict.SetContext(reg, restrict.ActorOwner).Allow(SpawnToolName) {
 		t.Fatal("owner was denied the spawn tool")
 	}