CVE-2026-78379

Consent bypass in python_repl tool via batch kwargs forwarding in Amazon Strands Agents Tools

Published Aug 25, 2026 CVSS 4.0 · 9.2 Critical EPSS < 1% KEV · Not listed
Description
Improper neutralization of input used for LLM prompting in the python_repl tool in Amazon Strands Agents Tools before 0.8.5 might allow remote actors to execute arbitrary Python code on the agent's host by bypassing the human consent gate, via a crafted prompt that forwards non_interactive_mode as a keyword argument through the batch tool. To remediate this issue, users should upgrade to version 0.8.5 or later.

Analysis

AI-generated

Technical impact

This vulnerability occurs in the python_repl tool of Amazon Strands Agents Tools. Prior to version 0.8.5, inputs used for large‑language‑model prompting are not properly neutralized. This flaw allows a remote actor to inject a crafted prompt that forwards the non_interactive_mode keyword argument to the batch tool, bypassing the human consent gate. A successful exploit can lead to arbitrary Python code execution on the host running the agent, compromising confidentiality, integrity, and availability of the system.

Affected Systems

Affected products are Amazon Strands Agents Tools, versions before 0.8.5. Systems that deploy the python_repl function are at risk unless they have upgraded to 0.8.5 or later.

Risk and Exploitability

The CVSS score of 9.2 signals a high‑severity remote code execution threat. EPSS data is not available, and the vulnerability is not listed in the CISA KEV catalog, so the current likelihood of exploitation is unknown. Based on the description, it is inferred that an attacker requires the ability to send prompts to the python_repl tool, which could be achieved through remote interfaces or interactions with the agent. Given the severity and the potential for unrestricted code execution, urgent review and mitigation are recommended.

Remediation

No solution or workaround provided in the CVE record.

OpenCVE Recommended Actions

  • Update Amazon Strands Agents Tools to version 0.8.5 or later.
  • If an immediate upgrade is not possible, disable or remove the python_repl component from the agent’s deployment, or enforce strict input validation to block non_interactive_mode forwarding.
  • Restrict network access to the agent and monitor for anomalous prompt activity on the python_repl endpoint.
  • Configure the agent to require explicit user confirmation before executing any LLM‑generated code.

Generated by OpenCVE AI on August 26, 2026 at 03:13 UTC.

Tracking

Sign in to view the affected projects.

Advisories

No advisories available
History
Type Values Removed Values Added
Metrics ssvc

{'options': {'Automatable': 'no', 'Exploitation': 'none', 'Technical Impact': 'total'}, 'version': '2.0.3'}

Type Values Removed Values Added
Description Improper neutralization of input used for LLM prompting in the python_repl tool in Amazon Strands Agents Tools before 0.8.5 might allow remote actors to execute arbitrary Python code on the agent's host by bypassing the human consent gate, via a crafted prompt that forwards non_interactive_mode as a keyword argument through the batch tool. To remediate this issue, users should upgrade to version 0.8.5 or later.
Title Consent bypass in python_repl tool via batch kwargs forwarding in Amazon Strands Agents Tools
First Time appeared Amazon
Amazon strands-agents-tools
Weaknesses CWE-1427
CPEs cpe:2.3:a:amazon:strands-agents-tools:*:*:*:*:*:*:*:*
Vendors & Products Amazon
Amazon strands-agents-tools
References
Metrics cvssV3_1

{'score': 8.1, 'vector': 'CVSS:3.1/AV:N/AC:H/PR:N/UI:N/S:U/C:H/I:H/A:H'}

cvssV4_0

{'score': 9.2, 'vector': 'CVSS:4.0/AV:N/AC:L/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:N/SI:N/SA:N'}

Weaknesses
  • CWE-1427

    Improper Neutralization of Input Used for LLM Prompting

Sources

Status
PUBLISHED
Assigner
AMZN
Published
Updated
Aug 25, 2026, 7:20 PM
Reserved
Aug 24, 2026, 1:19 PM
Status
Awaiting Analysis
Published
Aug 25, 2026, 7:16 PM
Modified
Aug 26, 2026, 4:45 PM