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/spawnpersonacontract_test.go

@@ -14,17 +14,14 @@ import (
 	"git.yeij.top/AskaEth/Cyrene_v2/internal/runtime/model"
 )
 
-// This file locks the CONTRACT half of design/22 §8-2 (the behavior half lives in
-// spawnpersona_test.go): the tool schema and result shape a model sees, the rejection
-// of an unknown role, and the session env a sub-agent run is stamped with. The point of
-// these tests is that a dispatch which names no persona stays byte-identical to the
-// pre-change behavior, while a dispatch that names an unusable role fails loudly rather
-// than degrading into the default role.
+// 本文件锁定 design/22 §8-2 的契约那一半（行为那一半在 spawnpersona_test.go）：模型
+// 看到的工具 schema 与结果形状、未知角色的拒绝，以及子代理运行被盖上的会话 env。这些
+// 测试的重点是：不指名 persona 的派发与改动前的行为逐字节一致，而指名一个不可用角色的
+// 派发会明确报错，而不是降级为默认角色。
 
-// TestSpawnPersonaRejectsUnknownRole is the "未知角色 → 明确报错，不静默回落" half: an id
-// the process cannot run is refused before anything starts, for a single dispatch and
-// for a batch (all-or-nothing, so no half-run batch is left behind), and a blank id is
-// never treated as a role.
+// TestSpawnPersonaRejectsUnknownRole 是「未知角色 → 明确报错，不静默回落」那一半：本
+// 进程跑不了的 id 会在任何东西启动之前就被拒绝，单次派发与批次都是如此（全有或全无，
+// 因此不会留下半跑的批次）；空 id 也绝不会被当作角色。
 func TestSpawnPersonaRejectsUnknownRole(t *testing.T) {
 	t.Parallel()
 	cases := []struct {
@@ -84,10 +81,9 @@ func TestSpawnPersonaRejectsUnknownRole(t *testing.T) {
 	}
 }
 
-// TestSchedulerSpawnCarriesPersonaEnv pins the library half of the same contract for
-// callers that bypass the tool: Spawn stamps the target role onto the run's env (so
-// the loop resolves that role's prompt and memory scope), a blank persona leaves the
-// inherited env untouched, and Spawn itself refuses an unknown role.
+// TestSchedulerSpawnCarriesPersonaEnv 为绕过该工具的调用方钉住同一契约的库那一半：
+// Spawn 把目标角色盖到运行的 env 上（使循环解析该角色的提示词与记忆作用域），空 persona
+// 让继承来的 env 原样不动，Spawn 自身也拒绝未知角色。
 func TestSchedulerSpawnCarriesPersonaEnv(t *testing.T) {
 	t.Parallel()
 	s := newScheduler(t, Options{PersonaExists: func(id string) bool { return id == targetRole }})
@@ -98,8 +94,8 @@ func TestSchedulerSpawnCarriesPersonaEnv(t *testing.T) {
 	if err != nil {
 		t.Fatalf("Spawn: %v", err)
 	}
-	// The recorded binding names the target role too, so the progress events and the
-	// ws subagent_event frame report the role the sub-agent answered as.
+	// 记录下来的绑定也点名目标角色，因此进度事件与 ws subagent_event 帧报告的是子代理
+	// 以之作答的角色。
 	if sub.RoleID != targetRole || sub.PersonaID != targetRole {
 		t.Fatalf("sub-agent binding = %q/%q, want the target role on both", sub.RoleID, sub.PersonaID)
 	}
@@ -126,9 +122,8 @@ func TestSchedulerSpawnCarriesPersonaEnv(t *testing.T) {
 	}
 }
 
-// envRunner records the session env each sub-agent run starts with. The memory scope
-// a run uses is decided from exactly this value, so it is the boundary under test for
-// callers that go through Scheduler.Spawn directly.
+// envRunner 记录每次子代理运行起步时的会话 env。一次运行使用哪个记忆作用域正是由这个值
+// 决定的，因此对直接走 Scheduler.Spawn 的调用方而言，它就是被测边界。
 type envRunner struct {
 	envs chan loop.SessionEnv
 }
@@ -138,10 +133,9 @@ func (r envRunner) Run(ctx context.Context, _ string, _ []model.Message) (*loop.
 	return &loop.Result{Text: "ok", Steps: 1}, nil
 }
 
-// TestSpawnToolResultShapeIsUnchangedWithoutPersona pins the strict backward
-// compatibility of the model-visible result: a dispatch that names no persona still
-// returns exactly the same JSON keys as before this feature — no "persona" field is
-// added anywhere — so nothing about the omitted-persona contract moved.
+// TestSpawnToolResultShapeIsUnchangedWithoutPersona 钉住模型可见结果的严格向后兼容：
+// 不指名 persona 的派发仍返回与引入本功能之前完全相同的 JSON 键 —— 任何地方都没有新增
+// "persona" 字段 —— 因此省略 persona 的契约一点没动。
 func TestSpawnToolResultShapeIsUnchangedWithoutPersona(t *testing.T) {
 	t.Parallel()
 	s, sp := spawnToolFor(t, Options{}, time.Second)
@@ -175,7 +169,7 @@ func TestSpawnToolResultShapeIsUnchangedWithoutPersona(t *testing.T) {
 	}
 }
 
-// sortedKeys returns m's keys in sorted order, for a stable comparison in tests.
+// sortedKeys 按字典序返回 m 的键，供测试做稳定比较。
 func sortedKeys(m map[string]any) []string {
 	out := make([]string, 0, len(m))
 	for k := range m {
@@ -185,17 +179,16 @@ func sortedKeys(m map[string]any) []string {
 	return out
 }
 
-// TestSpawnSchemaKeepsPersonaOptional pins the tool contract: persona is an optional
-// string at both the single-goal and the tasks level, nothing new is required, and the
-// pre-existing arguments keep their meaning — so a model (or an old client) that never
-// sends persona sees byte-identical validation behavior.
+// TestSpawnSchemaKeepsPersonaOptional 钉住工具契约：persona 在单 goal 与 tasks 两级都是
+// 可选字符串，没有新增任何必填项，既有参数含义不变 —— 因此从不发 persona 的模型（或旧
+// 客户端）看到的校验行为逐字节相同。
 func TestSpawnSchemaKeepsPersonaOptional(t *testing.T) {
 	t.Parallel()
 	props, err := NewSpawnTool(nil).Parameters()
 	if err != nil {
 		t.Fatalf("Parameters: %v", err)
 	}
-	// The required set is unchanged: goal or tasks, nothing else.
+	// 必填集合未变：goal 或 tasks，别无其他。
 	if got := requiredAnyOf(t, props); strings.Join(got, ",") != "goal,tasks" {
 		t.Fatalf("anyOf required = %v, want goal/tasks only", got)
 	}
@@ -216,7 +209,7 @@ func TestSpawnSchemaKeepsPersonaOptional(t *testing.T) {
 	}
 }
 
-// requiredAnyOf flattens the anyOf required lists of a tool schema.
+// requiredAnyOf 摊平工具 schema 的各个 anyOf required 列表。
 func requiredAnyOf(t *testing.T, props map[string]any) []string {
 	t.Helper()
 	anyOf, _ := props["anyOf"].([]any)
@@ -232,7 +225,7 @@ func requiredAnyOf(t *testing.T, props map[string]any) []string {
 	return out
 }
 
-// stringProp returns the named top-level property of a tool schema property block.
+// stringProp 返回工具 schema 属性块里指名的顶层属性。
 func stringProp(t *testing.T, props map[string]any, name string) map[string]any {
 	t.Helper()
 	properties, _ := props["properties"].(map[string]any)