diff --git a/VERSION b/VERSION
index fa18a7e81..4dc6dc76b 100644
--- a/VERSION
+++ b/VERSION
@@ -1 +1 @@
-4.8.6-SUPAI
+4.8.7-SUPAI
diff --git a/docs/release-readiness.md b/docs/release-readiness.md
index a332ec559..f8431152a 100644
--- a/docs/release-readiness.md
+++ b/docs/release-readiness.md
@@ -75,6 +75,7 @@ Decision owner:
 |---|---|---|---|---|---|
 | parser_v2 | off/shadow/enforce | off/shadow/enforce | hold/promote/disable | | |
 | context_engine | off/shadow/enforce | off/shadow/enforce | hold/promote/disable | | |
+| tool_calling | off/enforce | off/enforce | hold/promote/disable | | |
 | auto_continue | off/shadow/enforce | off/shadow/enforce | hold/promote/disable | | |
 | capability_router | off/shadow/enforce | off/shadow/enforce | hold/promote/disable | | |
 
@@ -131,6 +132,7 @@ Required follow-ups:
 |---|---|---|---|
 | `parser_v2` | shadow 入口已就绪；报告能统计 diff、confidence、marker leak；不改变对外响应 | diff 经人工审阅；marker leak 和 false positive 均有 fixtures；Go/Node parity 可解释 | 出现未解释 marker leak、tool call 丢失、跨 chunk 误判 |
 | `context_engine` | ContextPlan 和 warning 可记录；tool pair 风险可被报告引用；shadow 不改变请求 | 多轮任务 shadow 稳定；tool_call/tool_result 不断链；reasoning summary 不挤压当前请求 | tool pair orphan、当前请求被错误裁剪、current-input hash 异常 |
+| `tool_calling` | 全局关闭后能拒绝或剥离 tools；普通对话不受影响；输出侧不解析工具调用 | 默认开启路径单元测试通过；关闭/剥离路径有回归测试；WebUI / API 文档可回滚 | 携带 `tool_choice=auto` 的普通请求被误拒、关闭后仍注入 tools、关闭后仍解析工具调用 |
 | `auto_continue` | detector shadow 能记录 candidate、skip reason、trace；默认不续写 | non-stream smoke 通过；有 `max_continue_count`、`max_total_ms`、`max_extra_tokens`；失败可回退 | tool call、JSON mode、structured output、stream continuation 缺 live smoke |
 | `capability_router` | 所有公开 alias 有 profile；shadow 只产生 warning；冲突可被 History Analyzer 归类 | policy 经样本验证；冲突处理不破坏 search/thinking/current-input 语义 | unknown alias、search/thinking 冲突不可解释、vision/current-input 策略不完整 |
 
diff --git a/internal/readiness/baseline.go b/internal/readiness/baseline.go
index 6fbf824ff..419474b59 100644
--- a/internal/readiness/baseline.go
+++ b/internal/readiness/baseline.go
@@ -81,12 +81,28 @@ func defaultFeatureReadiness() []FeatureReadiness {
 			MissingEvidence: []string{"parser shadow report", "manual diff review"},
 		},
 		{
-			Feature:         "context_engine",
-			CurrentMode:     ModeOff,
-			TargetMode:      ModeShadow,
-			Decision:        FeatureHold,
-			Reason:          "waiting for context shadow evidence",
-			MissingEvidence: []string{"context shadow report", "tool pair review"},
+			Feature:     "context_engine",
+			CurrentMode: ModeEnforce,
+			TargetMode:  ModeEnforce,
+			Decision:    FeatureHold,
+			Reason:      "current default is enforce; keep rollback path and collect live evidence",
+			Evidence: []EvidenceRef{
+				{Source: "config defaults", Summary: "context_engine defaults to enforce with hybrid_recent"},
+				{Source: "offline current-input smoke", Ref: "./tests/scripts/run-offline-current-input-smoke.sh"},
+			},
+			MissingEvidence: []string{"live multi-turn account smoke", "context shadow report for future strategy changes"},
+		},
+		{
+			Feature:     "tool_calling",
+			CurrentMode: ModeEnforce,
+			TargetMode:  ModeEnforce,
+			Decision:    FeatureHold,
+			Reason:      "current default is enabled; settings switch provides reject or ignore_tools rollback",
+			Evidence: []EvidenceRef{
+				{Source: "unit tests", Ref: "./tests/scripts/run-unit-all.sh"},
+				{Source: "docs", Ref: "docs/toolcall-semantics.md"},
+			},
+			MissingEvidence: []string{"live tool-call account smoke"},
 		},
 		{
 			Feature:         "auto_continue",
diff --git a/internal/readiness/baseline_test.go b/internal/readiness/baseline_test.go
index 2d53cd4b7..abef49253 100644
--- a/internal/readiness/baseline_test.go
+++ b/internal/readiness/baseline_test.go
@@ -16,8 +16,26 @@ func TestBuildBaselineReportDefaultsPendingEvidence(t *testing.T) {
 	if report.Decision.Decision != DecisionGoFlagsOff {
 		t.Fatalf("decision = %q, want %q", report.Decision.Decision, DecisionGoFlagsOff)
 	}
-	if len(report.Features) != 4 {
-		t.Fatalf("features = %d, want 4", len(report.Features))
+	if len(report.Features) != 5 {
+		t.Fatalf("features = %d, want 5", len(report.Features))
+	}
+	var foundContext, foundToolCalling bool
+	for _, feature := range report.Features {
+		switch feature.Feature {
+		case "context_engine":
+			foundContext = true
+			if feature.CurrentMode != ModeEnforce || feature.TargetMode != ModeEnforce {
+				t.Fatalf("context_engine modes = %q/%q, want enforce/enforce", feature.CurrentMode, feature.TargetMode)
+			}
+		case "tool_calling":
+			foundToolCalling = true
+			if feature.CurrentMode != ModeEnforce || feature.TargetMode != ModeEnforce {
+				t.Fatalf("tool_calling modes = %q/%q, want enforce/enforce", feature.CurrentMode, feature.TargetMode)
+			}
+		}
+	}
+	if !foundContext || !foundToolCalling {
+		t.Fatalf("expected context_engine and tool_calling readiness rows, got %#v", report.Features)
 	}
 	if len(report.Shadow) != 5 {
 		t.Fatalf("shadow evidence = %d, want 5", len(report.Shadow))