CVE-2026-120497

AWS AI-DLC traceability reports can execute injected browser script

AWS · Former aidlc-traceability CLI

First documented disclosure · Review checkpoint

Summary

Untrusted project artifacts can alter a generated HTML report when someone opens it.

Severity and exploitation evidence

Medium · 4.6/10 · CVSS 4.0. Severity source.

Exploitation not established: Malicious exploitation is not established by the reviewed primary sources.

Exploitation evidence source

Technical severity is not a measure of observed attacks or of risk to your deployment. Attempts, demonstrations, and successful compromise retain the qualifications in the source.

How the weakness works

Artifact strings embedded inside a script block were not protected against closing-tag sequences.

Conditions required

Attacker-controlled artifact titles or descriptions, an unpatched CLI and a user opening its HTML output. AI mode and AWS credentials are unnecessary.

Potential impact

Script can read report contents and falsify displayed coverage. The advisory does not establish access to other local files or credentials.

Affected versions and configurations

From 0.1.0 before commit 96cf0eaa31a3ad8636e486c7fd2b91c55166393f.

Fix and mitigation status

Commit 96cf0ea, July 3; the CLI was removed from this repository July 16.

Practical next steps

Remove obsolete local copies or apply the July 3 fix, including in forks. Use Markdown output for untrusted projects until remediated.

Operating lesson

A generated assessment report is an active-content boundary, not inherently trusted evidence.

Evidence and limits

Public hardening PR 481 opened July 2; the fix merged July 3. AWS published the CVE and detailed advisory November 2. Adjacent file-read and symlink hardening in that PR is not automatically part of this CVE.

Sources