diff --git a/packages/ai/tools/handlers/figures.py b/packages/ai/tools/handlers/figures.py
index 0bf3a3e..cd6da0c 100644
--- a/packages/ai/tools/handlers/figures.py
+++ b/packages/ai/tools/handlers/figures.py
@@ -6,8 +6,6 @@
 from uuid import UUID
 
 from packages.ai.tools.types import ToolProgress, ToolResult
-from packages.storage.db import session_scope
-from packages.storage.repositories import PaperRepository
 
 if TYPE_CHECKING:
     from collections.abc import Iterator
@@ -16,25 +14,25 @@
 def _analyze_figures(paper_id: str, max_figures: int = 10) -> Iterator[ToolProgress | ToolResult]:
     """提取并解读论文图表"""
     from packages.ai.figure_service import FigureService
+    from packages.ai.tools.base import _require_paper
 
-    with session_scope() as session:
-        repo = PaperRepository(session)
-        try:
-            paper = repo.get_by_id(UUID(paper_id))
-        except (ValueError, Exception) as exc:
-            yield ToolResult(success=False, summary=f"论文不存在: {exc}")
-            return
-        if not paper.pdf_path:
-            yield ToolResult(success=False, summary="论文没有 PDF 文件，无法提取图表")
-            return
-        pdf_path = paper.pdf_path
-        title = paper.title
+    # 用 _require_paper 统一解析（支持短前缀，返回 detached 但属性已加载的 paper）
+    paper, err = _require_paper(paper_id)
+    if err:
+        yield err
+        return
+    if not paper.pdf_path:
+        yield ToolResult(success=False, summary="论文没有 PDF 文件，无法提取图表")
+        return
+    pdf_path = paper.pdf_path
+    title = paper.title
+    pid = UUID(paper.id)  # 用完整 UUID，不用原始短前缀
 
     yield ToolProgress(message=f"正在提取「{(title or '')[:30]}」中的图表...", current=1, total=3)
     svc = FigureService()
     try:
         results = svc.analyze_paper_figures(
-            UUID(paper_id),
+            pid,
             pdf_path,
             max_figures=max_figures,
         )
diff --git a/packages/ai/tools/handlers/read.py b/packages/ai/tools/handlers/read.py
index a3ed786..937475e 100644
--- a/packages/ai/tools/handlers/read.py
+++ b/packages/ai/tools/handlers/read.py
@@ -24,7 +24,9 @@ def _skim_paper(paper_id: str) -> Iterator[ToolProgress | ToolResult]:
     if not paper.abstract:
         yield ToolResult(success=False, summary="该论文缺少摘要，无法执行粗读")
         return
-    pid = UUID(paper_id)
+    # 用 paper.id（_require_paper 已 resolve 成完整 UUID），不用原始 paper_id
+    # （原始可能是短前缀，UUID(paper_id) 会崩 "badly formed hexadecimal UUID"）
+    pid = UUID(paper.id)
     title = (paper.title or "")[:40]
     yield ToolProgress(message=f"正在粗读「{title}」...", current=1, total=2)
     try:
@@ -48,7 +50,7 @@ def _deep_read_paper(paper_id: str) -> Iterator[ToolProgress | ToolResult]:
     if not paper.arxiv_id and not paper.pdf_path:
         yield ToolResult(success=False, summary="该论文无 arXiv ID 且无 PDF，无法精读")
         return
-    pid = UUID(paper_id)
+    pid = UUID(paper.id)
     title = (paper.title or "")[:40]
     yield ToolProgress(message=f"正在精读「{title}」，预计 30-60 秒...", current=1, total=3)
     try:
@@ -68,11 +70,11 @@ def _embed_paper(paper_id: str) -> Iterator[ToolProgress | ToolResult]:
     if err:
         yield err
         return
-    pid = UUID(paper_id)
+    pid = UUID(paper.id)
     if paper.embedding:
         yield ToolResult(
             success=True,
-            data={"paper_id": paper_id, "status": "already_embedded"},
+            data={"paper_id": paper.id, "status": "already_embedded"},
             summary="该论文已有向量，跳过",
         )
         return
@@ -87,7 +89,7 @@ def _embed_paper(paper_id: str) -> Iterator[ToolProgress | ToolResult]:
         PaperPipelines().embed_paper(pid)
         yield ToolResult(
             success=True,
-            data={"paper_id": paper_id, "status": "embedded"},
+            data={"paper_id": paper.id, "status": "embedded"},
             summary="向量化完成",
         )
     except Exception as exc:
diff --git a/packages/ai/tools/handlers/reasoning.py b/packages/ai/tools/handlers/reasoning.py
index ced908f..4cb7ee2 100644
--- a/packages/ai/tools/handlers/reasoning.py
+++ b/packages/ai/tools/handlers/reasoning.py
@@ -6,8 +6,6 @@
 from uuid import UUID
 
 from packages.ai.tools.types import ToolProgress, ToolResult
-from packages.storage.db import session_scope
-from packages.storage.repositories import PaperRepository
 
 if TYPE_CHECKING:
     from collections.abc import Iterator
@@ -16,20 +14,20 @@
 def _reasoning_analysis(paper_id: str) -> Iterator[ToolProgress | ToolResult]:
     """推理链深度分析"""
     from packages.ai.reasoning_service import ReasoningService
+    from packages.ai.tools.base import _require_paper
 
-    with session_scope() as session:
-        repo = PaperRepository(session)
-        try:
-            paper = repo.get_by_id(UUID(paper_id))
-        except (ValueError, Exception) as exc:
-            yield ToolResult(success=False, summary=f"论文不存在: {exc}")
-            return
-        title = paper.title
+    # 用 _require_paper 统一解析（支持短前缀，返回 detached 但属性已加载的 paper）
+    paper, err = _require_paper(paper_id)
+    if err:
+        yield err
+        return
+    title = paper.title
+    pid = UUID(paper.id)  # 用完整 UUID，不用原始短前缀
 
     yield ToolProgress(message=f"正在分析「{(title or '')[:30]}」的推理链...", current=1, total=2)
     svc = ReasoningService()
     try:
-        result = svc.analyze(UUID(paper_id))
+        result = svc.analyze(pid)
     except Exception as exc:
         yield ToolResult(success=False, summary=f"推理链分析失败: {exc}")
         return