diff --git a/src/core/pipeline/runner.py b/src/core/pipeline/runner.py
index 4999b71..6b46269 100644
--- a/src/core/pipeline/runner.py
+++ b/src/core/pipeline/runner.py
@@ -67,8 +67,18 @@ PIPELINE_STEPS: List[StepSpec] = [
     ),
     StepSpec(
         step_id="step5", method_name="step5_extract_training_spectra",
-        requires=["deglint_img_path", "csv_path", "boundary_path", "glint_mask_path"],
+        requires=["deglint_img_path", "processed_csv_path", "csv_path", "boundary_path", "glint_mask_path"],
         produces=["training_csv_path"],
+        # processed_csv_path(step4 产物) 才是 step5 真正需要的主路径，
+        # 通过 parameter_map 显式映射到形参 csv_path。
+        # raw csv_path 也保留在 requires 中以备 user_config 覆盖，
+        # 但用占位名 _raw_csv_ignored 注入，落到 step5 形参列表末尾的 **kwargs 兜底。
+        # 这样可以避免 L2 顺序注入中"后注入的 csv_path=None 覆盖前面的 processed_csv_path"的冲突。
+        parameter_map={
+            "processed_csv_path": "csv_path",
+            "csv_path": "_raw_csv_ignored",
+        },
+        skip_when_missing=False,
         description="实测样本点光谱提取",
     ),
     StepSpec(
@@ -157,9 +167,10 @@ class PipelineRunner:
                 missing = [k for k in spec.requires if not ctx.get(k)]
                 if missing:
                     ctx.status[spec.step_id] = "skipped"
-                    ctx.append_log(
-                        f"[RUNNER] {spec.step_id} 缺少入参: {missing}，跳过"
-                    )
+                    reason = f"缺少必要的上下文参数，自动跳过: {missing}"
+                    ctx.append_log(f"[RUNNER] {spec.step_id} {reason}")
+                    if hasattr(self.pipeline, "_notify"):
+                        self.pipeline._notify(spec.description, "skipped", reason)
                     continue
             self._invoke(spec, ctx)
         ctx.pipeline_end_time = time.time()