// Security Advisory Triage
Docling arbitrary file read/write via untrusted TikZ rendering
GitHub advisories — pip · published 2026-10-08 13:35 UTC
WHAT CHANGEDCVE-2026-105744 published 7 October 2026. Docling and docling-slim before 2.132.0 allow arbitrary file read/write when rendering untrusted TikZ with the opt-in Tectonic engine. Command execution is possible if shell-escape is enabled. Patched in 2.132.0.
WHY IT MATTERSServices that process untrusted documents with Docling's Tectonic engine can be exploited to read or write arbitrary files on the host. If shell-escape is enabled, attackers gain command execution. Document processing pipelines that accept user uploads are the primary attack surface.
WHAT TO DOGrep your dependency trees for docling and docling-slim. If found and you use the Tectonic engine for rendering, update to 2.132.0 immediately. If you cannot update, disable Tectonic rendering or block TikZ content at ingestion.
Model-reported confidence 85% (not measured accuracy)
Primary source ↗
written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+ghsa_id: GHSA-x3q2-h9hx-4r4j +summary: Docling has arbitrary file read/write (and command execution when shell-escape is enabled) when rendering untrusted TikZ with the opt-in Tectonic engine +severity: high +published_at: 2026-10-07T20:42:05Z +cve_id: CVE-2026-105744 +vulnerabilities[].package.name: docling, docling-slim +vulnerabilities[].first_patched_version: 2.132.0, 2.132.0
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
| triage verdict | MATERIAL at 0.78, needed 0.5 |
|---|---|
| triage reasoning | High-severity arbitrary file read/write and RCE in document processing library with available patch. |
| signals | known exploitation not stated, but high severity + arbitrary file read/write + command execution · docling is a document processing library likely in dependency trees of services handling untrusted input · new advisory with CVE assignment and patched version available · attack surface (rendering untrusted TikZ) is plausible in document processing workflows |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |