ZeroHour

CVE-2026-27966

PoC moderate

Prompt-Injection RCE in Langflow CSV Agent before 1.8.0

CVSS 3.1
9.8 critical
EPSS
34%p98
Published
()
Modified
AI analysis

Langflow versions prior to 1.8.0 hardcode allow_dangerous_code=True in the CSV Agent node, which silently exposes LangChain's Python REPL tool (python_repl_ast) to the agent. An attacker who can influence prompts or data processed by a flow using this node, for example via chat input or uploaded CSV content, can inject instructions that make the REPL execute arbitrary Python and, from there, OS commands. This yields full remote code execution on the server with the privileges of the Langflow process, enabling data theft, tampering, and further compromise. Any deployment of Langflow below 1.8.0 whose flows use the CSV Agent node is affected, with internet-facing or multi-user instances at highest risk. No confirmed in-the-wild exploitation is reported yet (not in CISA KEV), but a public security advisory exists and EPSS assigns a 33.7% probability of exploitation within 30 days, so patching should be treated as urgent.

What to do: Upgrade Langflow to version 1.8.0 or later. Until patched, avoid the CSV Agent node in any flow that processes untrusted prompts or files, and restrict network exposure of Langflow instances rather than leaving them reachable without authentication. Check server logs for unexpected Python or shell command execution originating from agent flows.

Affected
langflowAll versions prior to 1.8.0 (exploitation requires flows using the CSV Agent node); fixed in 1.8.0
Estimated exposure
moderatelikely on the order of tens of thousands of self-hosted Langflow deployments, of which a smaller subset (thousands) is internet-exposed — Estimate based on Langflow's large open-source community and adoption as an AI agent builder, combined with prior internet-wide scans that have found tens of thousands of exposed Langflow servers; the vulnerable code path additionally…

Order-of-magnitude estimate by the model from install counts, market share and public scan data it knows; verify before quoting.

Description

Langflow is a tool for building and deploying AI-powered agents and workflows. Prior to version 1.8.0, the CSV Agent node in Langflow hardcodes `allow_dangerous_code=True`, which automatically exposes LangChain’s Python REPL tool (`python_repl_ast`). As a result, an attacker can execute arbitrary Python and OS commands on the server via prompt injection, leading to full Remote Code Execution (RCE). Version 1.8.0 fixes the issue.

Vendors
langflow
Products
langflow
Weakness
CWE-94
Vector
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H

In the news