diff --git "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/02-\346\236\266\346\236\204/02-NoName-\344\273\243\347\220\206\347\263\273\347\273\237.md" "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/02-\346\236\266\346\236\204/02-NoName-\344\273\243\347\220\206\347\263\273\347\273\237.md"
index 7724e71..0764871 100644
--- "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/02-\346\236\266\346\236\204/02-NoName-\344\273\243\347\220\206\347\263\273\347\273\237.md"
+++ "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/02-\346\236\266\346\236\204/02-NoName-\344\273\243\347\220\206\347\263\273\347\273\237.md"
@@ -73,7 +73,7 @@
 
 ### 5.1 命令层压力正在下降
 
-`reviewed apply` 的请求构造、快照校验、显式 apply、人工 review intent 与 second guardrail 记录逻辑已经下沉到 `noname_apply.rs`。
+`reviewed apply` 的 request input 构造、快照校验、显式 apply、人工复核结果写回、review intent、second guardrail 预校验与记录逻辑已经下沉到 `noname_apply.rs`。
 
 `tauri_commands.rs` 在这条链路上主要保留：
 
@@ -82,7 +82,7 @@
 - game engine / plot state 读取与回写
 - 命令错误返回
 
-后续仍可继续观察低风险 apply 与生成主链中是否还有适合抽离的重复编排，但不应再把新的 reviewed apply 权限逻辑直接堆回命令层。
+这意味着 reviewed apply 的命令层已经进一步变成薄壳，本轮 A2/A3 可视为阶段性收口完成。后续仍可继续观察低风险 apply 与生成主链中是否还有适合抽离的重复编排，但不应再把新的 reviewed apply 权限逻辑直接堆回命令层。
 
 ### 5.2 Web mock 与后端容易漂移
 
diff --git "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/01-\347\216\260\347\212\266\350\257\204\344\274\260.md" "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/01-\347\216\260\347\212\266\350\257\204\344\274\260.md"
index a3f4f8b..d63975b 100644
--- "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/01-\347\216\260\347\212\266\350\257\204\344\274\260.md"
+++ "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/01-\347\216\260\347\212\266\350\257\204\344\274\260.md"
@@ -24,13 +24,14 @@
 
 `tauri_commands.rs` 仍然承担大量主游戏编排工作，继续向里面叠逻辑会提高维护风险和冲突概率。
 
-当前 `NoName reviewed apply` 链路已经完成第三轮减压：
+当前 `NoName reviewed apply` 链路已经完成第五轮减压：
 
 - reviewed apply request 构造、快照校验与显式 apply 已下沉到 `noname_apply.rs`
 - trace 获取 / plot state 更新 / trace 回写已复用命令层共用壳层
-- human review intent 与 second guardrail 决策记录已下沉到 `noname_apply.rs`
+- human review intent、人工复核结果写回与 second guardrail 预校验 / 决策记录已下沉到 `noname_apply.rs`
+- `apply_noname_reviewed_output` 的 request input 组装也已回收到 `noname_apply.rs`
 
-后续风险重点不再是 reviewed apply 本身，而是其他低风险 apply 与剧情生成主链是否继续膨胀。
+因此 A2/A3 在当前主开发线可以视为阶段性完成。后续风险重点不再是 reviewed apply 本身，而是其他低风险 apply 与剧情生成主链是否继续膨胀。
 
 ### 3.2 前后端 mock 语义漂移
 
diff --git "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/02-\345\220\216\347\273\255\350\267\257\347\272\277\345\233\276.md" "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/02-\345\220\216\347\273\255\350\267\257\347\272\277\345\233\276.md"
index ade044e..0167de4 100644
--- "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/02-\345\220\216\347\273\255\350\267\257\347\272\277\345\233\276.md"
+++ "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/02-\345\220\216\347\273\255\350\267\257\347\272\277\345\233\276.md"
@@ -53,12 +53,13 @@
 
 当前状态：
 
-- 已完成第三轮下沉
+- 已完成第五轮下沉
 - reviewed apply 的 request builder 与 plotText manual apply helper 已进入 `noname_apply.rs`
 - `tauri_commands.rs` 在 reviewed apply 路径上已减少内联快照构造与包装逻辑
 - reviewed apply 命令已开始复用共享的 trace 读取、plot state 更新和 trace 回写壳层
-- human review intent 与 second guardrail 决策记录已进入 `noname_apply.rs`
-- 下一步若继续减压，应优先评估低风险 apply 或剧情生成主链中仍重复的编排，而不是扩大新的 apply scope
+- human review intent、人工复核结果写回与 second guardrail 预校验 / 决策记录已进入 `noname_apply.rs`
+- `apply_noname_reviewed_output` 的 request input 组装已回收到 `noname_apply.rs`
+- 本轮后 A2/A3 可视为阶段性完成；下一步若继续减压，应优先评估低风险 apply 或剧情生成主链中仍重复的编排，而不是扩大新的 apply scope
 
 ## P2：收口后继续推进
 
diff --git "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/04-\346\234\252\346\235\245\344\270\244\345\221\250\345\274\200\345\217\221\346\216\222\346\234\237.md" "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/04-\346\234\252\346\235\245\344\270\244\345\221\250\345\274\200\345\217\221\346\216\222\346\234\237.md"
index 52c5c54..e1d119d 100644
--- "a/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/04-\346\234\252\346\235\245\344\270\244\345\221\250\345\274\200\345\217\221\346\216\222\346\234\237.md"
+++ "b/docs/\351\241\271\347\233\256\346\226\207\346\241\243/03-\350\247\204\345\210\222/04-\346\234\252\346\235\245\344\270\244\345\221\250\345\274\200\345\217\221\346\216\222\346\234\237.md"
@@ -91,11 +91,12 @@
 
 当前状态：
 
-- 第三轮已完成
+- 第五轮已完成
 - `NoNameReviewedApplyRequest` 构造与 plotText manual apply helper 已从命令层下沉到 `noname_apply.rs`
 - trace 获取 / result 回写的共用壳层已开始抽出并复用到两个 reviewed apply 命令
-- human review intent 与 second guardrail 决策记录已从命令层下沉到 `noname_apply.rs`
-- 下一轮若继续 A2，应只评估剩余低风险 apply / 生成主链重复编排，不新增 apply scope，不扩张 second guardrail 语义
+- human review intent、人工复核结果写回与 second guardrail 预校验 / 决策记录已从命令层下沉到 `noname_apply.rs`
+- `apply_noname_reviewed_output` 的 request input 组装已回收到 `noname_apply.rs`
+- 当前 A2 在主开发线已阶段性完成；后续若继续 A2，应只评估剩余低风险 apply / 生成主链重复编排，不新增 apply scope，不扩张 second guardrail 语义
 
 ### 任务 A3：建立 `NoName` 主链变更的文档闭环
 
@@ -113,6 +114,12 @@
 - 最小验证：
   - 文档自查，无失效路径引用
 
+当前状态：
+
+- `NoName` 架构、现状评估、后续路线图已同步 reviewed apply 最新收口进度
+- 两周排期文档已记录 A2 的第五轮下沉结果，后续 review 不再依赖口头补充
+- 后续若继续推进主开发线，应维持“改主链实现就同步三份主文档和排期状态”的节奏
+
 ## 4.2 协作者线
 
 ### 任务 B1：记忆检索排序解释性增强
diff --git a/src-tauri/src/noname_apply.rs b/src-tauri/src/noname_apply.rs
index bc7f060..35d85ce 100644
--- a/src-tauri/src/noname_apply.rs
+++ b/src-tauri/src/noname_apply.rs
@@ -1,3 +1,4 @@
+use crate::noname_output_interface::NoNameControlledOutputDecision;
 use crate::noname_trace::{NoNameHumanReviewDecision, NoNameSecondGuardrailDecision, NoNameTrace};
 use crate::noname_types::{
     NoNameApplyScope, NoNameMode, NoNameProposal, NoNameProposalStatus, NoNameTargetSegment,
@@ -171,6 +172,13 @@ fn next_apply_plan_order(trace: &NoNameTrace) -> u32 {
         + 1
 }
 
+fn find_controlled_output_review_index(trace: &NoNameTrace, request_id: &str) -> Option<usize> {
+    trace
+        .controlled_output_reviews
+        .iter()
+        .position(|review| review.request_id == request_id)
+}
+
 fn find_review_proposal<'a>(
     trace: &'a NoNameTrace,
     request_id: &str,
@@ -316,6 +324,45 @@ fn human_review_apply_intent_rejection_reason(
     }
 }
 
+pub fn apply_human_review_decision_to_trace(
+    trace: &mut NoNameTrace,
+    request_id: &str,
+    decision: NoNameHumanReviewDecision,
+    reviewed_at: u64,
+) -> Result<(), String> {
+    let review_index = find_controlled_output_review_index(trace, request_id)
+        .ok_or_else(|| format!("NoName controlled output review not found: {}", request_id))?;
+
+    let safe_apply_scope = {
+        let review = &trace.controlled_output_reviews[review_index];
+        if review.decision != NoNameControlledOutputDecision::NeedsReview
+            || !review.requires_human_review
+        {
+            return Err(format!(
+                "NoName controlled output review does not require human review: {}",
+                request_id
+            ));
+        }
+
+        review.safe_apply_scope.ok_or_else(|| {
+            format!(
+                "NoName controlled output review has no safe apply scope: {}",
+                request_id
+            )
+        })?
+    };
+
+    let review = &mut trace.controlled_output_reviews[review_index];
+    review.human_review_decision = Some(decision);
+    review.human_reviewed_at = Some(reviewed_at);
+    review.human_review_note = Some(decision.note().to_string());
+
+    trace.record_proposal_transition(format!("{}:human_review:{}", request_id, decision.as_str()));
+    record_human_review_apply_intent(trace, request_id, decision, Some(safe_apply_scope));
+
+    Ok(())
+}
+
 pub fn record_human_review_apply_intent(
     trace: &mut NoNameTrace,
     request_id: &str,
@@ -502,6 +549,32 @@ pub fn record_second_guardrail_decision(
     Ok(())
 }
 
+pub fn resolve_second_guardrail_for_trace(
+    trace: &mut NoNameTrace,
+    request_id: &str,
+    decision: NoNameSecondGuardrailDecision,
+) -> Result<(), String> {
+    let safe_apply_scope = trace
+        .controlled_output_reviews
+        .iter()
+        .find(|review| review.request_id == request_id)
+        .ok_or_else(|| format!("NoName controlled output review not found: {}", request_id))
+        .and_then(|review| {
+            if review.human_review_decision
+                != Some(NoNameHumanReviewDecision::ApprovedForHigherApply)
+            {
+                return Err(format!(
+                    "NoName controlled output review is not approved for second guardrail: {}",
+                    request_id
+                ));
+            }
+
+            Ok(review.safe_apply_scope)
+        })?;
+
+    record_second_guardrail_decision(trace, request_id, decision, safe_apply_scope)
+}
+
 fn trace_has_second_guardrail_allow(
     trace: &NoNameTrace,
     request_id: &str,
@@ -909,6 +982,15 @@ pub fn apply_reviewed_output_to_plot_state(
     }
 }
 
+pub fn apply_reviewed_output_input_to_plot_state(
+    trace: NoNameTrace,
+    input: NoNameReviewedApplyRequestInput,
+    plot_state: PlotState,
+) -> Result<NoNameReviewedApplyOutcome, String> {
+    let request = build_reviewed_apply_request(input)?;
+    apply_reviewed_output_to_plot_state(trace, request, plot_state)
+}
+
 pub fn apply_manual_plot_text_hint_to_plot_state(
     trace: NoNameTrace,
     request_id: &str,
@@ -1140,6 +1222,67 @@ mod tests {
             .any(|entry| entry.ends_with(":apply_intent:awaiting_second_guardrail")));
     }
 
+    #[test]
+    fn apply_human_review_decision_to_trace_updates_review_and_logs() {
+        let (mut trace, request_id) = make_trace_for_plot_text_review();
+
+        apply_human_review_decision_to_trace(
+            &mut trace,
+            &request_id,
+            NoNameHumanReviewDecision::ApprovedForHigherApply,
+            123,
+        )
+        .expect("human review decision should be applied to trace");
+
+        let review = &trace.controlled_output_reviews[0];
+        assert_eq!(
+            review.human_review_decision,
+            Some(NoNameHumanReviewDecision::ApprovedForHigherApply)
+        );
+        assert_eq!(review.human_reviewed_at, Some(123));
+        assert_eq!(
+            review.human_review_note.as_deref(),
+            Some(NoNameHumanReviewDecision::ApprovedForHigherApply.note())
+        );
+        assert!(trace.proposal_transition_log.iter().any(|entry| {
+            entry
+                == &format!(
+                    "{}:human_review:{}",
+                    request_id,
+                    NoNameHumanReviewDecision::ApprovedForHigherApply.as_str()
+                )
+        }));
+        assert!(trace.apply_execution_log.iter().any(|entry| {
+            entry.target == "plot_text_hint" && entry.outcome == "awaiting_second_guardrail"
+        }));
+    }
+
+    #[test]
+    fn apply_human_review_decision_to_trace_rejects_missing_safe_scope_without_mutation() {
+        let (mut trace, request_id) = make_trace_for_plot_text_review();
+        trace.controlled_output_reviews[0].safe_apply_scope = None;
+
+        let error = apply_human_review_decision_to_trace(
+            &mut trace,
+            &request_id,
+            NoNameHumanReviewDecision::ApprovedForHigherApply,
+            456,
+        )
+        .expect_err("missing safe apply scope should fail");
+
+        assert_eq!(
+            error,
+            format!(
+                "NoName controlled output review has no safe apply scope: {}",
+                request_id
+            )
+        );
+        let review = &trace.controlled_output_reviews[0];
+        assert_eq!(review.human_review_decision, None);
+        assert_eq!(review.human_reviewed_at, None);
+        assert_eq!(review.human_review_note, None);
+    }
+
     #[test]
     fn human_review_apply_intent_rejects_non_second_guardrail_scope() {
         let (mut trace, request_id) = make_trace_for_summary_apply();
@@ -1190,6 +1333,49 @@ mod tests {
         );
     }
 
+    #[test]
+    fn resolve_second_guardrail_for_trace_requires_review_approval() {
+        let (mut trace, request_id) = make_trace_for_plot_text_review();
+
+        let error = resolve_second_guardrail_for_trace(
+            &mut trace,
+            &request_id,
+            NoNameSecondGuardrailDecision::Allow,
+        )
+        .expect_err("second guardrail should require approved human review");
+
+        assert_eq!(
+            error,
+            format!(
+                "NoName controlled output review is not approved for second guardrail: {}",
+                request_id
+            )
+        );
+    }
+
+    #[test]
+    fn resolve_second_guardrail_for_trace_records_allow() {
+        let (mut trace, request_id) = make_trace_for_plot_text_review();
+        apply_human_review_decision_to_trace(
+            &mut trace,
+            &request_id,
+            NoNameHumanReviewDecision::ApprovedForHigherApply,
+            789,
+        )
+        .expect("human review should prepare second guardrail");
+
+        resolve_second_guardrail_for_trace(
+            &mut trace,
+            &request_id,
+            NoNameSecondGuardrailDecision::Allow,
+        )
+        .expect("second guardrail allow should be recorded");
+
+        assert!(trace.apply_execution_log.iter().any(|entry| {
+            entry.target == "plot_text_hint" && entry.outcome == "second_guardrail_allowed"
+        }));
+    }
+
     #[test]
     fn second_guardrail_allows_plot_augmentation_hint() {
         let (mut trace, request_id) = make_trace_for_plot_augmentation_apply();
@@ -1283,6 +1469,34 @@ mod tests {
             .any(|entry| entry.ends_with(":manual_apply:chapter_summary_hint")));
     }
 
+    #[test]
+    fn reviewed_apply_input_can_write_chapter_summary_hint() {
+        let (trace, request_id) = make_trace_for_summary_apply();
+        let plot_state = make_plot_state("existing summary");
+
+        let outcome = apply_reviewed_output_input_to_plot_state(
+            trace,
+            NoNameReviewedApplyRequestInput {
+                request_id,
+                scope: NoNameApplyScope::ChapterSummaryHint,
+                chapter_index: Some(1),
+                segment_index: None,
+                expected_segment_text: None,
+                expected_summary: Some("existing summary".to_string()),
+                expected_generation_diagnostics: None,
+                expected_plot_augmentation_hints: None,
+            },
+            plot_state,
+        )
+        .expect("summary hint should be manually applied from request input");
+
+        assert!(outcome
+            .plot_state
+            .current_chapter
+            .summary
+            .starts_with("NoName summary hint: sect crisis; existing summary"));
+    }
+
     #[test]
     fn reviewed_apply_rejects_stale_chapter_summary_snapshot() {
         let (trace, request_id) = make_trace_for_summary_apply();
diff --git a/src-tauri/src/tauri_commands.rs b/src-tauri/src/tauri_commands.rs
index b8aa8dd..ab78fa5 100644
--- a/src-tauri/src/tauri_commands.rs
+++ b/src-tauri/src/tauri_commands.rs
@@ -15,16 +15,15 @@ use crate::llm_runtime_config::{
 use crate::llm_service::{LLMConfig, LLMRequest, LLMService};
 use crate::memory_layers::{ChapterSummary, MemoryEntry, WorldFact};
 use crate::noname_apply::{
-    apply_manual_plot_text_hint_to_plot_state, apply_reviewed_output_to_plot_state,
-    build_reviewed_apply_request, record_human_review_apply_intent,
-    record_second_guardrail_decision, NoNameReviewedApplyRequestInput,
+    apply_human_review_decision_to_trace, apply_manual_plot_text_hint_to_plot_state,
+    apply_reviewed_output_input_to_plot_state, resolve_second_guardrail_for_trace,
+    NoNameReviewedApplyRequestInput,
 };
 use crate::noname_config::NoNameConfig;
 use crate::noname_context_builder::build_context_packet;
 use crate::noname_context_types::NoNameContextBuildInput;
 use crate::noname_guardrails::{NoNameDirectorGuardrailInput, NoNameGuardrailResult};
 use crate::noname_memory_manager::NoNameMemoryManager;
-use crate::noname_output_interface::NoNameControlledOutputDecision;
 use crate::noname_roles::NoNameDirectorObservation;
 use crate::noname_runtime::{NoNameDirectorRunResult, NoNameRuntime, NoNameTurnInput};
 use crate::noname_trace::{NoNameHumanReviewDecision, NoNameSecondGuardrailDecision, NoNameTrace};
@@ -4250,37 +4249,7 @@ pub async fn mark_noname_controlled_output_review(
         .duration_since(UNIX_EPOCH)
         .map(|duration| duration.as_secs())
         .unwrap_or_default();
-    let safe_apply_scope = {
-        let review = trace
-            .controlled_output_reviews
-            .iter_mut()
-            .find(|review| review.request_id == request_id)
-            .ok_or_else(|| format!("NoName controlled output review not found: {}", request_id))?;
-
-        if review.decision != NoNameControlledOutputDecision::NeedsReview
-            || !review.requires_human_review
-        {
-            return Err(format!(
-                "NoName controlled output review does not require human review: {}",
-                request_id
-            ));
-        }
-
-        review.human_review_decision = Some(decision);
-        review.human_reviewed_at = Some(reviewed_at);
-        review.human_review_note = Some(decision.note().to_string());
-        review.safe_apply_scope
-    };
-
-    if safe_apply_scope.is_none() {
-        return Err(format!(
-            "NoName controlled output review has no safe apply scope: {}",
-            request_id
-        ));
-    }
-
-    trace.record_proposal_transition(format!("{}:human_review:{}", request_id, decision.as_str()));
-    record_human_review_apply_intent(&mut trace, &request_id, decision, safe_apply_scope);
+    apply_human_review_decision_to_trace(&mut trace, &request_id, decision, reviewed_at)?;
 
     if !runtime.replace_trace(trace.clone()) {
         return Err(format!("NoName trace not found: {}", trace_id));
@@ -4305,24 +4274,7 @@ pub async fn resolve_noname_second_guardrail(
         .find(|trace| trace.trace_id == trace_id)
         .ok_or_else(|| format!("NoName trace not found: {}", trace_id))?;
 
-    let safe_apply_scope = {
-        let review = trace
-            .controlled_output_reviews
-            .iter()
-            .find(|review| review.request_id == request_id)
-            .ok_or_else(|| format!("NoName controlled output review not found: {}", request_id))?;
-
-        if review.human_review_decision != Some(NoNameHumanReviewDecision::ApprovedForHigherApply) {
-            return Err(format!(
-                "NoName controlled output review is not approved for second guardrail: {}",
-                request_id
-            ));
-        }
-
-        review.safe_apply_scope
-    };
-
-    record_second_guardrail_decision(&mut trace, &request_id, decision, safe_apply_scope)?;
+    resolve_second_guardrail_for_trace(&mut trace, &request_id, decision)?;
 
     if !runtime.replace_trace(trace.clone()) {
         return Err(format!("NoName trace not found: {}", trace_id));
@@ -4366,18 +4318,21 @@ pub async fn apply_noname_reviewed_output(
     expected_plot_augmentation_hints: Option<Vec<String>>,
     engine: State<'_, Mutex<GameEngine>>,
 ) -> Result<NoNameManualPlotTextApplyResult, String> {
-    let request = build_reviewed_apply_request(NoNameReviewedApplyRequestInput {
-        request_id,
-        scope,
-        chapter_index,
-        segment_index,
-        expected_segment_text,
-        expected_summary,
-        expected_generation_diagnostics,
-        expected_plot_augmentation_hints,
-    })?;
     apply_noname_outcome_with_engine(&trace_id, &engine, |trace, plot_state| {
-        apply_reviewed_output_to_plot_state(trace, request, plot_state)
+        apply_reviewed_output_input_to_plot_state(
+            trace,
+            NoNameReviewedApplyRequestInput {
+                request_id,
+                scope,
+                chapter_index,
+                segment_index,
+                expected_segment_text,
+                expected_summary,
+                expected_generation_diagnostics,
+                expected_plot_augmentation_hints,
+            },
+            plot_state,
+        )
     })
 }
 
@@ -5082,6 +5037,7 @@ mod tests {
     use super::*;
     use crate::event_log::{EventImportance, GameEvent};
     use crate::models::{CultivationRealm, Element, Grade, SpiritualRoot};
+    use crate::noname_output_interface::NoNameControlledOutputDecision;
     use crate::script::{InitialState, Location, ScriptType, WorldSetting};
     use std::sync::{Mutex as TestMutex, OnceLock as TestOnceLock};
     use tempfile::tempdir;