Disclosed vulnerability
CVE-2026-31020: server-side template injection to RCE in DocsGPT custom-prompt feature
AIR-2026-0220 AIR record · 2026-09-docsgpt-custom-prompt-ssti-rce
From the supporting source
An unauthenticated attacker can inject malicious template expressions, leading to a server-side template injection (SSTI) vulnerability that can be exploited to achieve full remote code execution (RCE).cveawg.mitre.org
The record
Disclosed
2026-09-04
Vendor
Arc53 / DocsGPT
Product
DocsGPT
Risk domain
Access control and permissions › Privilege escalation
System type
Generative model
Who acted
AI was the target — template injection in the agent framework; no model decision
Attack surface
Agent framework
Autonomy level
Assistant
Initial vector
Unauthenticated network request
Impact type
Remote code execution
Confidence
High
Submitted by
EAI Internal
Crosswalk
- OWASP Agentic Top 10 (2026)
- ASI03 – Identity and Privilege Abuse
- MITRE ATLAS
- AML.T0055 – Unsecured Credentials
- NIST AI RMF
- GOVERN 6.1
- EU AI Act
- Article 13
- ISO/IEC
- TR 24028; 42001; 23894
- AIUC-1
- B. Security
- Scope
- tool misuse
- Mitigation
- Apply least-privilege principle; Regular policy audits.
- Taxonomy
- Enkrypt AI agent risk taxonomy · Black-Box Red Teaming of Agentic AI: A Taxonomy-Driven Framework for Automated Risk Discovery
Blast radius
DocsGPT 0.15.0 and earlier
Capability abused
Taxonomy & references
Curator notes
CVSS 9.8. Published 2026-09-04. The agentic surface is the LLM custom-prompt feature, whose Jinja rendering is unsandboxed. PoC/disclosure; no real victim.
Supporting sources
Cite as
Enkrypt AI. "CVE-2026-31020: server-side template injection to RCE in DocsGPT custom-prompt feature". AIR, AIR-2026-0220. https://agentincident.vercel.app/i/AIR-2026-0220/ (accessed 2026-10-19).