All findings

CVE-2026-84973 high

GitHub's AI-agent index adds Jupyter and DataDog workflow findings

GitHub Security Lab's official AI-agent index covers a high-severity Jupyter checkout race with CVE-2026-84973 and an accepted DataDog fork workflow-execution issue without a CVE.

Bug class
Checkout time-of-check/time-of-use race and privileged workflow code execution
Affected codebase
JupyterLab maintainer-tools and DataDog's Trivy fork CI workflows
Credited system
GitHub Security Lab AI agent (unspecified)
Latest disclosure
September 4, 2026
Coverage
Project acceptance August 19-September 4, 2026
Attribution
Direct source attribution
AI role
Unspecified
Severity
high
Source status: GitHub Security Lab lists GHSL-2026-203 and GHSL-2026-199 in its official AI-agent index without naming a particular agent. Project or GHSL records establish acceptance; only the Jupyter finding has a CVE.

Summary

GHSL-2026-203 describes a checkout race in JupyterLab maintainer-tools. The project advisory maps it to CVE-2026-84973, rates it high, and identifies 1.0.0 as fixed. The public CVE API had not yet synchronized at the audit cutoff, but the maintainer GHSA supplies the exact identifier and accepted fix.

GHSL-2026-199 covers privileged workflow execution in DataDog’s fork of Trivy. The project accepted the report and deleted the affected workflow on August 19. It has no CVE and concerns the fork’s CI workflow, not a product vulnerability in the upstream Trivy scanner.

GitHub’s official index establishes AI-agent provenance for both advisories but does not identify Taskflow or another particular agent. Bugflation therefore uses the existing system-unspecified profile.


References


Catalogued in the Bugflation public ledger. Disagree with the attribution or severity label? Email the desk.