// Security Advisory Triage

CVE-2026-78379 - Consent bypass in Strands Agents Tools python_repl tool

AWS security bulletins · published 2026-10-22 17:18 UTC
ACTION NEEDED VERIFY_EXPOSURE SECURITY
WHAT CHANGEDCVE-2026-78379 - Consent bypass in Strands Agents Tools python_repl tool. Source event: 2026-10-21T19:54:11+00:00. RSS discovery: 2026-10-22T04:46:21.247402+00:00.
WHY IT MATTERSThe captured feed identifies CVE-2026-78379. Check exposure against the affected product and versions; detailed remediation evidence is unavailable.
WHAT TO DOCheck your inventory for the product and affected versions named in this advisory. Remediation details were not captured; consult the vendor bulletin before changing configuration or applying a fix. Remediation: see bulletin https://aws.amazon.com/security/security-bulletins/2026-089-aws/ (retrieval attempted 2026-10-22T17:18:01.615Z; RSS discovery 2026-10-22T04:46:21.247402+00:00).
Deterministic source evidence; confidence not scored Primary source ↗ written by advisory-evidence-v2
Show the reasoning — raw diff, triage verdict, confidence
RSS DISCOVERY DIFF
 ## CVE-2026-78379 - Consent bypass in Strands Agents Tools python_repl tool
 https://aws.amazon.com/security/security-bulletins/rss/2026-089-aws/
-Wed, 09 Sep 2026 21:30:11 +0000
+Wed, 21 Oct 2026 19:54:11 +0000
 
-Bulletin ID: 2026-089-AWS
-
-Scope: AWS
-
-Content Type: Important (requires attention)
-
-Publication Date: 08/25/2026 12:00 PM PDT
+Bulletin ID: 2026-089-AWS Scope: AWS Content Type: Important (requires attention) Publication Date: 08/25/2026 12:00 PM PDT
 Description:
 Strands Agents is an open-source Python SDK for building and running AI agents. The strands-agents-tools package provides pre-built tools for use with the SDK, including the python_repl tool, which executes Python code on the agent's host, and the batch tool, which invokes several other tools in a single call. Before executing code, python_repl prompts the operator for approval. We identified CVE-2026-78379, a consent bypass in the python_repl tool. 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.
 Impacted versions: < 0.8.5
SEPARATE BULLETIN CAPTURE · unavailable

Retrieved separately from the RSS discovery; this does not establish historical page content.

retrieval attempted2026-10-22T17:18:01.615Z
captured atunavailable
sourcehttps://aws.amazon.com/security/security-bulletins/2026-089-aws/
content SHA-256unavailable
bulletin retrieval or evidence validation failed
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
triage verdictMATERIAL at 0.92, needed 0.55
triage reasoningRemote code execution in AI agent SDK via consent bypass. Affects Python services running Strands Agents.
signalsAI SDK advisory · consent bypass / authentication bypass · arbitrary code execution on host · remote exploitation vector · fixed version available (0.8.5) · open-source Python package in dependency tree · bulletin re-dated (metadata update only, but advisory itself is current)
suppression—
triage modelanthropic/claude-haiku-4.5 ($0.00000)
writer modeladvisory-evidence-v2 ($0.00000)