diff --git a/app/processor/base.py b/app/processor/base.py
index c1a2996..6700caf 100644
--- a/app/processor/base.py
+++ b/app/processor/base.py
@@ -55,9 +55,9 @@ class BaseNode(ABC):
         """子类实现节点核心逻辑，返回状态更新"""
 
     def fail_on_invalid_output(self, state: T, error: ValueError, reason: str) -> T:
-        """模型输出未通过校验时收尾为失败结果，详细原因只写日志"""
+        """模型输出未通过校验时收尾为失败结果，原因交给失败收尾由模型向用户说明"""
         self.logger.warning(f"{reason}: {error}")
-        return {**state, "status": "failed", "failure_reason": reason}
+        return {**state, "status": "failed", "failure_reason": f"{reason}：{error}"}
 
     def log_step(self, step_name: str, message: str = ""):
         """记录步骤日志。"""
diff --git a/app/processor/graph.py b/app/processor/graph.py
index 9e766d1..807a9c7 100644
--- a/app/processor/graph.py
+++ b/app/processor/graph.py
@@ -248,8 +248,8 @@ def build_graph(checkpointer=None):
         },
     )
 
-    # 失败节点统一写入最终失败状态后结束流程
-    workflow.add_edge("fallback_node", END)
+    # 失败节点写入最终失败状态后，由回答节点向用户说明原因
+    workflow.add_edge("fallback_node", "qa_node")
 
     # 用户补充后重新合并输入并判断意图，取消则直接结束
     workflow.add_conditional_edges(
diff --git a/app/processor/nodes/action_node.py b/app/processor/nodes/action_node.py
index 5a598ab..979c203 100644
--- a/app/processor/nodes/action_node.py
+++ b/app/processor/nodes/action_node.py
@@ -192,6 +192,16 @@ class ActionNode(BaseNode):
             updated_state["validation_errors"] = [GENERATION_ERROR]
             return updated_state
 
+        # 模型确认现有能力无法满足需求时直接失败收尾，不进入重试
+        if batch.unsupported_reason:
+            updated_state.update(
+                actions=[],
+                validation_passed=False,
+                failure_reason=f"动作生成判断无法完成：{batch.unsupported_reason}",
+                status="failed",
+            )
+            return updated_state
+
         # 保存生成的动作，实际目标、规划一致性及模拟执行结果仍由后续校验节点检查
         updated_state["actions"] = batch.actions
 
diff --git a/app/processor/nodes/fallback_node.py b/app/processor/nodes/fallback_node.py
index fac2a43..0b2dbaf 100644
--- a/app/processor/nodes/fallback_node.py
+++ b/app/processor/nodes/fallback_node.py
@@ -1,31 +1,40 @@
-"""统一收尾能力不足、资源缺失和动作修正失败"""
+"""统一收尾能力不足、资源缺失和动作修正失败
+
+这里只整理内部失败原因，面向用户的说明由回答节点交给模型总结，代码中不写死给用户的文案
+"""
 
 from app.processor.base import BaseNode
 from app.processor.state import QueryState
 
 
+def failure_notice(state: QueryState) -> str:
+    """汇总上游记录的失败原因、最近一次修正分析和校验错误，作为交给模型的事实说明"""
+    facts = []
+    if reason := state.get("failure_reason"):
+        facts.append(reason)
+    plan = state.get("repair_plan")
+    if plan is not None and plan.reason not in facts:
+        facts.append(f"失败分析：{plan.reason}")
+    if errors := state.get("validation_errors"):
+        facts.append("校验错误：" + "；".join(error.message for error in errors))
+        facts.append(f"已自动修正 {state.get('retry_count', 0)} 次")
+    detail = "\n".join(facts) if facts else "没有记录具体原因"
+    return f"本次修改请求没有执行，内部原因如下：\n{detail}"
+
+
 class FallbackNode(BaseNode):
     name = "fallback_node"
 
     def process(self, state: QueryState) -> QueryState:
         # 浅拷贝状态
         updated_state: QueryState = state.copy()
-        # 读取上游记录的具体失败原因
-        reason = state.get("failure_reason")
+        notice = failure_notice(state)
+        self.logger.warning(notice)
 
-        # 如果上游没有提供有效原因，则根据校验错误补充失败说明
-        if not reason:
-            # 如果仍有校验错误，使用达到自动修正上限的说明，并带上已重试次数
-            if state.get("validation_errors"):
-                reason = f"动作校验未通过，已达到自动修正上限，共重试 {state.get('retry_count', 0)} 次"
-            else:
-                # 如果既没有具体原因也没有校验错误，使用通用说明，避免返回空原因
-                reason = "当前需求未能完成，缺少具体失败原因"
-
-        # 将任务标记为最终失败，结束本轮处理
+        # 将任务标记为最终失败，由回答节点向用户说明原因
         updated_state["status"] = "failed"
-        # 保存上游原因或本节点补充的说明，供后续输出给用户
-        updated_state["failure_reason"] = reason
+        updated_state["failure_reason"] = notice
+        updated_state["notice"] = notice
         # 显式标记校验未通过，防止残留动作被当作可执行结果
         updated_state["validation_passed"] = False
         # 清空旧模拟快照，避免失败结果中仍携带之前的成功快照
diff --git a/app/processor/nodes/planner_node.py b/app/processor/nodes/planner_node.py
index 1ce2e24..bff5496 100644
--- a/app/processor/nodes/planner_node.py
+++ b/app/processor/nodes/planner_node.py
@@ -65,7 +65,7 @@ class PlannerNode(BaseNode):
                 # 解析失败、引用不存在的目标等可由模型自行纠正，超过次数才失败收尾
                 if attempt == PLAN_MAX_REPAIRS:
                     return self.fail_on_invalid_output(
-                        state, error, "未能生成可用的大屏规划，请调整需求后重试"
+                        state, error, "规划输出多次未通过校验"
                     )
                 self.logger.warning(f"规划未通过校验，重新规划: {error}")
                 feedback = f"上一次输出未通过校验：{error}\n请修正后重新输出完整规划。"
diff --git a/app/processor/nodes/qa_node.py b/app/processor/nodes/qa_node.py
index 9aae076..6e5384f 100644
--- a/app/processor/nodes/qa_node.py
+++ b/app/processor/nodes/qa_node.py
@@ -38,7 +38,7 @@ class QANode(BaseNode):
 
         content = state["input_context"].model_dump_json(exclude_unset=True)
         offer = None
-        # 落地策略判定请求无法执行时，附上事实说明，由本节点如实告知用户
+        # 请求无法执行或执行失败时，附上事实说明，由本节点如实告知用户
         if notice := state.get("notice"):
             content += f"\n\n系统说明：{notice}"
             # 回复中的替代修改需要结构化保存，下一轮用户同意时直接执行，因此不流式输出
@@ -59,8 +59,9 @@ class QANode(BaseNode):
         updated_state["answer"] = answer
         # 保存等待用户确认的提议，没有提议时清空
         updated_state["offer"] = offer
-        # 回答生成完成，问答分支在此结束
-        updated_state["status"] = "completed"
+        # 回答生成完成；失败收尾后的说明保持失败状态，不交付动作
+        if state.get("status") != "failed":
+            updated_state["status"] = "completed"
 
         # 返回包含回答和完成状态的更新结果
         return updated_state
diff --git a/app/processor/nodes/reflection_node.py b/app/processor/nodes/reflection_node.py
index fa778c9..f3fa01e 100644
--- a/app/processor/nodes/reflection_node.py
+++ b/app/processor/nodes/reflection_node.py
@@ -15,7 +15,6 @@ from app.processor.state import QueryState
 from app.prompt.action import FIELD_QUERY_PROMPT, uses_material_fields
 from app.prompt.reflection import REFLECTION_SYSTEM_PROMPT
 from app.schemas.action_repair import ActionRepairPlan
-from app.schemas.capabilities import USER_FAILURE_MESSAGE
 
 
 class ReflectionNode(BaseNode):
@@ -88,11 +87,9 @@ class ReflectionNode(BaseNode):
         updated_state["repair_plan"] = plan
         # 仅在建议允许重试时将次数加一，缺少历史计数时从零开始，不可重试时保持原次数
         updated_state["retry_count"] = state.get("retry_count", 0) + int(plan.retryable)
-        # 可以重试时清空旧失败原因；不可重试时建议中的分析含内部术语，只记录日志，对用户使用统一说明
-        if not plan.retryable:
-            self.logger.warning("动作无法修正: {}", plan.reason)
+        # 可以重试时清空旧失败原因；不可重试时保留分析，由失败收尾交给模型向用户说明
         updated_state["failure_reason"] = (
-            None if plan.retryable else USER_FAILURE_MESSAGE
+            None if plan.retryable else f"修正分析判断无法完成：{plan.reason}"
         )
 
         # 返回修正建议、重试次数和失败原因，由后续路由决定重新生成动作还是结束
diff --git a/app/processor/nodes/requirement_node.py b/app/processor/nodes/requirement_node.py
index b57278a..709a8a1 100644
--- a/app/processor/nodes/requirement_node.py
+++ b/app/processor/nodes/requirement_node.py
@@ -18,9 +18,7 @@ class RequirementNode(BaseNode):
             return await self._generate(state)
         except ValueError as error:
             # 结构化输出解析失败或引用校验不通过（含 ValidationError、OutputParserException）
-            return self.fail_on_invalid_output(
-                state, error, "未能可靠理解你的需求，请换个说法再试"
-            )
+            return self.fail_on_invalid_output(state, error, "需求理解的输出未通过校验")
 
     async def _generate(self, state: QueryState) -> QueryState:
         # 浅拷贝状态
diff --git a/app/processor/nodes/scope_node.py b/app/processor/nodes/scope_node.py
index 03023bd..b90a347 100644
--- a/app/processor/nodes/scope_node.py
+++ b/app/processor/nodes/scope_node.py
@@ -22,9 +22,7 @@ class ScopeNode(BaseNode):
             return await self._generate(state)
         except ValueError as error:
             # 结构化输出解析失败或引用校验不通过（含 ValidationError、OutputParserException）
-            return self.fail_on_invalid_output(
-                state, error, "未能确定需要修改的范围，请描述得更具体一些"
-            )
+            return self.fail_on_invalid_output(state, error, "修改范围的输出未通过校验")
 
     async def _generate(self, state: QueryState) -> QueryState:
         # 浅拷贝状态
diff --git a/app/prompt/action.py b/app/prompt/action.py
index 92a16d7..dfb7cbc 100644
--- a/app/prompt/action.py
+++ b/app/prompt/action.py
@@ -12,6 +12,7 @@ ACTION_SYSTEM_PROMPT = f"""# 任务目标
 # 输出要求
 按提供的 ActionBatch Schema 返回 actions 列表，仅使用本次列出的动作类型
 覆盖本次全部变更，不返回大屏 JSON、解释文本或执行成功声明
+有任何变更无法用列出的动作或可用字段完成时，不执行其中一部分：actions 为空，在 unsupported_reason 中说明缺少什么
 以 input_context.screen 为起点，后续动作基于前序结果计算，不能重复执行或遗漏目标
 
 # 输入边界
@@ -58,7 +59,7 @@ field_groups 按物料类型列出本次相关组件的字段分组，节点的
 确定要改的字段：用用户原话中对属性的说法（input_context 中的提问，不用改写后的需求）调用 search_fields；
 返回多个字段时，依据字段说明判断用户指的是哪一个；用户看的是大屏画面，没有提到图层列表、属性面板等编辑器位置时，按画面上看得到的内容理解
 需要了解一组相关字段时用 get_fields，一次可查多个分组
-只能使用查询到的字段路径，值须符合字段的类型、可选值和取值范围；查不到对应字段说明不支持，不自行编造路径
+只能使用查询到的字段路径，值须符合字段的类型、可选值和取值范围；查不到对应字段说明物料不支持，按无法完成处理，不自行编造路径
 """
 
 # 这些动作会写组件属性，需要先查物料字段
diff --git a/app/prompt/qa.py b/app/prompt/qa.py
index 0c1d2cb..20756c5 100644
--- a/app/prompt/qa.py
+++ b/app/prompt/qa.py
@@ -19,8 +19,9 @@ QA_SYSTEM_PROMPT = """## 任务
 可以简短提示还能继续帮忙，不列举能力清单
 
 ## 无法执行的请求
-输入附带“系统说明”时，表示用户的修改请求因目标不存在而没有执行
-用一两句话如实说明原因，不声称已修改
+输入附带“系统说明”时，表示用户的修改请求没有执行，系统说明记录了内部原因
+用一两句话如实说明原因，不声称已修改；用用户能理解的话转述，不提字段路径、节点 ID、错误代码、重试次数等内部信息
+原因是当前能力不支持时直接说明做不到什么；原因是生成或校验出错时说明这次没能完成，可以换个说法再试
 有明显可行的替代修改时，只提一个，answer 以是否需要执行的问句结尾，例如“需要我新增一张折线图吗？”；不提供配置建议等非修改类替代
 同时把该替代修改写成 offer：一条独立完整、可直接执行的修改指令，例如“在当前大屏新增一张折线图”；没有替代修改时 offer 为 null
 新增组件使用物料示例数据，不要求用户提供数据；不要求用户离开对话去选中或操作组件
diff --git a/app/resources/materials/area-chart.json b/app/resources/materials/area-chart.json
index fe447ed..666d080 100644
--- a/app/resources/materials/area-chart.json
+++ b/app/resources/materials/area-chart.json
@@ -42,6 +42,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/bar-chart.json b/app/resources/materials/bar-chart.json
index 6d02315..55a1242 100644
--- a/app/resources/materials/bar-chart.json
+++ b/app/resources/materials/bar-chart.json
@@ -118,6 +118,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/bar-line-chart.json b/app/resources/materials/bar-line-chart.json
index b6361ae..b52b8f3 100644
--- a/app/resources/materials/bar-line-chart.json
+++ b/app/resources/materials/bar-line-chart.json
@@ -143,6 +143,7 @@
         }
       ]
     },
+    "chart-dataset",
     {
       "key": "bar-line-data",
       "label": "数据映射",
diff --git a/app/resources/materials/field_groups.json b/app/resources/materials/field_groups.json
index 099a15c..dd247d5 100644
--- a/app/resources/materials/field_groups.json
+++ b/app/resources/materials/field_groups.json
@@ -293,6 +293,28 @@
       }
     ]
   },
+  {
+    "key": "chart-dataset",
+    "label": "图表数据",
+    "description": "图表画面上展示的数据，柱子高度、折线点、扇区大小等数值都来自这里",
+    "fields": [
+      {
+        "path": "props.option.dataset.source",
+        "label": "数据",
+        "description": "图表展示的数据表，每个对象是一行（一个类目或数据项），键是列名，列名与数据映射字段对应；修改时提交完整的新数组：保留不需要修改的行和列，只改用户提到的数值或行；新增列或改列名时同时修改数据映射字段",
+        "value_type": "array",
+        "item_type": "object",
+        "aliases": [
+          "数据",
+          "图表数据",
+          "数据值",
+          "示例数据",
+          "数据集"
+        ],
+        "panel": false
+      }
+    ]
+  },
   {
     "key": "chart-encode-xy",
     "label": "数据映射",
diff --git a/app/resources/materials/funnel-chart.json b/app/resources/materials/funnel-chart.json
index e303ea8..860dc6f 100644
--- a/app/resources/materials/funnel-chart.json
+++ b/app/resources/materials/funnel-chart.json
@@ -69,6 +69,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/gauge-chart.json b/app/resources/materials/gauge-chart.json
index dccc2eb..b9b310f 100644
--- a/app/resources/materials/gauge-chart.json
+++ b/app/resources/materials/gauge-chart.json
@@ -82,6 +82,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/hbar-chart.json b/app/resources/materials/hbar-chart.json
index bf8eae6..38931ad 100644
--- a/app/resources/materials/hbar-chart.json
+++ b/app/resources/materials/hbar-chart.json
@@ -168,6 +168,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/heatmap-chart.json b/app/resources/materials/heatmap-chart.json
index cf74de4..348096e 100644
--- a/app/resources/materials/heatmap-chart.json
+++ b/app/resources/materials/heatmap-chart.json
@@ -137,6 +137,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy",
     {
       "key": "heatmap-value",
diff --git a/app/resources/materials/line-chart.json b/app/resources/materials/line-chart.json
index 87b23b6..0a75f1b 100644
--- a/app/resources/materials/line-chart.json
+++ b/app/resources/materials/line-chart.json
@@ -24,6 +24,7 @@
     "chart-grid",
     "chart-axis",
     "chart-line-series",
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/pictorial-bar-chart.json b/app/resources/materials/pictorial-bar-chart.json
index 3ef60bc..59fbe98 100644
--- a/app/resources/materials/pictorial-bar-chart.json
+++ b/app/resources/materials/pictorial-bar-chart.json
@@ -124,6 +124,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/pie-chart.json b/app/resources/materials/pie-chart.json
index 9130080..2efd65e 100644
--- a/app/resources/materials/pie-chart.json
+++ b/app/resources/materials/pie-chart.json
@@ -23,6 +23,7 @@
     "chart-legend",
     "chart-palette",
     "chart-item-series",
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/progress-ring-chart.json b/app/resources/materials/progress-ring-chart.json
index 70ced2e..12c77f0 100644
--- a/app/resources/materials/progress-ring-chart.json
+++ b/app/resources/materials/progress-ring-chart.json
@@ -65,6 +65,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/radar-chart.json b/app/resources/materials/radar-chart.json
index 1032793..d04d7ac 100644
--- a/app/resources/materials/radar-chart.json
+++ b/app/resources/materials/radar-chart.json
@@ -102,6 +102,7 @@
         }
       ]
     },
+    "chart-dataset",
     {
       "key": "radar-data",
       "label": "数据映射",
diff --git a/app/resources/materials/ring-chart.json b/app/resources/materials/ring-chart.json
index 90778b1..d5e11e1 100644
--- a/app/resources/materials/ring-chart.json
+++ b/app/resources/materials/ring-chart.json
@@ -43,6 +43,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/rose-chart.json b/app/resources/materials/rose-chart.json
index 1f82950..2b90653 100644
--- a/app/resources/materials/rose-chart.json
+++ b/app/resources/materials/rose-chart.json
@@ -64,6 +64,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-item"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/scatter-chart.json b/app/resources/materials/scatter-chart.json
index 082eb07..c65b743 100644
--- a/app/resources/materials/scatter-chart.json
+++ b/app/resources/materials/scatter-chart.json
@@ -67,6 +67,7 @@
         }
       ]
     },
+    "chart-dataset",
     "chart-encode-xy"
   ],
   "eventOptions": [
diff --git a/app/resources/materials/stacked-bar-chart.json b/app/resources/materials/stacked-bar-chart.json
index de15880..ff6a21e 100644
--- a/app/resources/materials/stacked-bar-chart.json
+++ b/app/resources/materials/stacked-bar-chart.json
@@ -89,6 +89,7 @@
         }
       ]
     },
+    "chart-dataset",
     {
       "key": "stacked-data",
       "label": "数据映射",
diff --git a/app/schemas/actions/__init__.py b/app/schemas/actions/__init__.py
index db6b701..bf27f44 100644
--- a/app/schemas/actions/__init__.py
+++ b/app/schemas/actions/__init__.py
@@ -1,8 +1,8 @@
 """动作类型统一入口，新增操作时在这里显式加入联合类型"""
 
-from typing import Annotated
+from typing import Annotated, Self
 
-from pydantic import Field, TypeAdapter
+from pydantic import Field, TypeAdapter, model_validator
 
 from .add import AddAction
 from .base import ActionModel, BaseAction
@@ -22,6 +22,16 @@ action_adapter = TypeAdapter(Action)
 
 class ActionBatch(ActionModel):
     actions: list[Action] = Field(description="按数组顺序执行的动作列表，允许为空")
+    unsupported_reason: str | None = Field(
+        default=None,
+        description="需求中有任何部分无法用现有动作或查到的字段完成时，说明缺少什么能力，此时 actions 为空；能完成时为 null",
+    )
+
+    @model_validator(mode="after")
+    def unsupported_has_no_actions(self) -> Self:
+        if self.unsupported_reason and self.actions:
+            raise ValueError("填写 unsupported_reason 时 actions 必须为空，不能只执行需求的一部分")
+        return self
 
 
 __all__ = [
diff --git a/app/schemas/agent.py b/app/schemas/agent.py
index 03c4b5b..f4e2eed 100644
--- a/app/schemas/agent.py
+++ b/app/schemas/agent.py
@@ -21,6 +21,5 @@ class AgentResult(BaseModel):
     message_id: str  # 本次用户输入的 ID，不是助手回复 ID
     status: Literal["completed", "waiting_for_input", "failed", "cancelled"]
     actions: list[Action] = Field(default_factory=list)
-    answer: str | None = None
+    answer: str | None = None  # 回答，或失败时由模型总结的原因说明
     clarification: Clarification | None = None
-    failure_reason: str | None = None
diff --git a/app/schemas/capabilities.py b/app/schemas/capabilities.py
index 5e08d01..f64841d 100644
--- a/app/schemas/capabilities.py
+++ b/app/schemas/capabilities.py
@@ -20,6 +20,3 @@ RESOURCE_QUERY_SUPPORTED = False
 RESOURCE_FALLBACK_ASSUMPTION = (
     "尚未接入物料目录之外的资源查询，所需内容使用物料默认配置和示例数据"
 )
-
-# 面向用户的失败说明，不暴露内部校验术语
-USER_FAILURE_MESSAGE = "抱歉，这次没能完成修改，请换个说法或稍后再试"
diff --git a/app/services/agent.py b/app/services/agent.py
index c97bfa4..41bd1d2 100644
--- a/app/services/agent.py
+++ b/app/services/agent.py
@@ -158,7 +158,7 @@ class AgentService:
         if result.status == "cancelled":
             return "用户取消了本次补充"
         if result.status == "failed":
-            return _clip(f"处理失败：{result.failure_reason or '未知原因'}")
+            return _clip(f"处理失败：{result.answer or '未知原因'}")
         if result.answer:
             return _clip(result.answer)
         requirement = state.get("requirement")
@@ -176,8 +176,10 @@ class AgentService:
                     prompt=state["request"]["prompt"],
                     outcome=result.status,
                     summary=self._summarize(state, result),
-                    # 只有正常结束的回复中的提议才等待用户确认
-                    offer=state.get("offer") if result.status == "completed" else None,
+                    # 回答或失败说明中的提议等待用户确认，等待补充和取消时没有提议
+                    offer=state.get("offer")
+                    if result.status in ("completed", "failed")
+                    else None,
                     created_at=datetime.now(timezone.utc),
                 )
             )
@@ -198,7 +200,6 @@ class AgentService:
             actions=state["actions"] if deliver else [],
             answer=state.get("answer"),
             clarification=state.get("clarification") if waiting else None,
-            failure_reason=state.get("failure_reason"),
         )
 
     async def stream(self, thread_id, payload, session_id, user_id=None):