// Security Advisory Triage
AWS security-agent-mcp-server 0.1.1–0.2.0 allows arbitrary file write via argument injection
AWS security bulletins · published 2026-10-02 20:05 UTC
WHAT CHANGEDAWS published CVE-2026-97662 on 1 October 2026. security-agent-mcp-server versions >= 0.1.1 and < 0.2.0 contain an argument injection flaw in the diff scan operation. A crafted reference value is interpreted as a command-line option, allowing arbitrary file creation, overwrite, or truncation outside the workspace directory, bypassing workspace confinement.
WHY IT MATTERSAn attacker who can control the reference parameter passed to a diff scan can write files anywhere on the host filesystem the server process can reach. This breaks the workspace isolation that prevents AI assistants from modifying system files or other projects. If your AI tooling calls this MCP server, an adversary with prompt injection or context control can escalate to host-level file manipulation.
WHAT TO DOCheck whether security-agent-mcp-server is in your dependency tree or running in your AI assistant infrastructure. If you are on 0.1.1 through 0.1.x, upgrade to 0.2.0 or later. If you cannot upgrade immediately, disable the diff scan feature or restrict what values can be passed as the reference parameter until you patch.
Model-reported confidence 95% (not measured accuracy)
Primary source ↗
written by anthropic/claude-sonnet-4.5
Show the reasoning — raw diff, triage verdict, confidence
RAW DIFF
+## CVE-2026-97662 - Argument injection in AWS security-agent-mcp-server diff scan +https://aws.amazon.com/security/security-bulletins/rss/2026-121-aws/ +Thu, 01 Oct 2026 18:16:35 +0000 + +Bulletin ID: 2026-121-AWS + +Scope: AWS + +Content Type: Important (requires attention) + +Publication Date: 10/01/2026 10:30 AM PDT +Description: +security-agent-mcp-server is an open-source Model Context Protocol (MCP) server, published by AWS in the awslabs/mcp repository, that AI assistants use to run local security scans (including differential "diff" scans) over source code. We identified CVE-2026-97662, an argument injection issue in the diff scan operation: a crafted reference value supplied to the scan is interpreted as a command-line option rather than a revision, which lets a context-dependent actor create, overwrite, or truncate arbitrary files on the host outside the intended workspace directory, bypassing the server's workspace-confinement control. +Impacted versions: >= 0.1.1 AND < 0.2.0 +Please refer to the article below for the most up-to-date and complete information related to this AWS Security Bulletin.
HOW IT WAS JUDGED
Triage signals below are model-reported assertions, preserved verbatim; they are not independent arithmetic verification.
| triage verdict | MATERIAL at 0.95, needed 0.55 |
|---|---|
| triage reasoning | AWS MCP server is an AI SDK/framework tool that security leads running AI agents would have in their stack. Argument injection enabling arbitrary file writes is exploitable. |
| signals | known exploitation vector (argument injection) · affects AI provider SDK/framework (AWS MCP server) · arbitrary file write outside workspace · security-critical control bypass (workspace confinement) · open-source tool likely in dependency trees of AI-integrated services · new CVE with fixed version range specified |
| suppression | — |
| triage model | anthropic/claude-haiku-4.5 ($0.00000) |
| writer model | anthropic/claude-sonnet-4.5 ($0.00000) |