{"id":2203,"cve":"CVE-2026-69255","cve_url":"https:\/\/www.cvedetails.com\/cve\/CVE-2026-69255\/","cwe":"CWE-94","cwe_number":94,"cwe_url":"https:\/\/cwe.mitre.org\/data\/definitions\/94","cvss_score":9.2,"published":"2026-08-04","owasp":"A03 Injection","odc":"Checking","cdc":"Missing Check","category":"Input Validation and Sanitization","subcategory":"Command Injection","scope":"Remote","impact":"Arbitrary Code Execution","summary":"Flowise CSVAgent allows RCE via Python code injection in CSV files.","permalink":"https:\/\/netpack.pt\/vaitp\/vulnerability\/CVE-2026-69255\/","description":"Flowise is a drag & drop user interface to build a customized large language model flow. Prior to 3.1.3, the CSVAgent in packages\/components\/nodes\/agents\/CSVAgent\/CSVAgent.ts extracted attacker-controlled CSV data with file.split(',').pop() and interpolated it directly into executable Python as base64_string = \"${base64String}\" before calling Pyodide. The validatePythonCodeForDataFrame() denylist only checked later LLM-generated code and did not validate this initial code block. An authenticated attacker could inject a closing quote followed by Python code, use Pyodide's js bridge to load Node.js child_process, and execute arbitrary operating system commands as root in the Flowise container. This issue is fixed in version 3.1.3.","solution":"Upgrade to Flowise version 3.1.3 or later.","affected_component":"Flowise","fixed_by_upgrading":true,"vulnerable_code":"class CSVAgent:\n    def _get_pyodide_code(self, file_content: str) -> str:\n        # Simulate extracting the base64 data from a data URI.\n        # The real vulnerability was in TS: file.split(',').pop()\n        try:\n            base64_string = file_content.split(',')[1]\n        except IndexError:\n            base64_string = \"\"\n\n        # The validatePythonCodeForDataFrame() function was not called on this\n        # initial code, only on later LLM-generated code.\n\n        # VULNERABLE: The extracted string is directly interpolated into the Python code block,\n        # allowing an attacker to inject code by closing the string literal.\n        python_code = f\"\"\"\nimport pandas as pd\nimport base64\nfrom io import StringIO\n\nbase64_string = \"{base64_string}\"\ndecoded_data = base64.b64decode(base64_string)\ncsv_file = StringIO(decoded_data.decode('utf-8'))\ndf = pd.read_csv(csv_file)\n# ... more code to process df would follow\n\"\"\"\n        return python_code\n","patched_code":"import json\n\nclass CSVAgent:\n    def _get_pyodide_code(self, file_content: str) -> str:\n        # Simulate extracting the base64 data from a data URI.\n        # The real vulnerability was in TS: file.split(',').pop()\n        try:\n            base64_string = file_content.split(',')[1]\n        except IndexError:\n            base64_string = \"\"\n\n        # The validatePythonCodeForDataFrame() function was not called on this\n        # initial code, only on later LLM-generated code.\n\n        # FIX: The string is serialized using json.dumps() to ensure it is treated\n        # as a single, safe string literal within the generated Python code.\n        python_code = f\"\"\"\nimport pandas as pd\nimport base64\nfrom io import StringIO\n\nbase64_string = {json.dumps(base64_string)}\ndecoded_data = base64.b64decode(base64_string)\ncsv_file = StringIO(decoded_data.decode('utf-8'))\ndf = pd.read_csv(csv_file)\n# ... more code to process df would follow\n\"\"\"\n        return python_code\n","payload_needed":true,"payload":"```\ndata:text\/csv;base64,,\"\\nimport js\\njs.require('child_process').execSync('touch \/tmp\/pwned')\\n#\n```","payload_status":"available","published_fix":[],"references":null}