
StepJack
Benchmarking framework for evaluating computer-use AI agents against multi-step indirect prompt injection, with automatic adversarial goal…

Benchmarking framework for evaluating computer-use AI agents against multi-step indirect prompt injection, with automatic adversarial goal…

Conducting cybersecurity testing using the pi agent in a pre-built sandbox environment.

Disabled TLS Certificate Verification for HashiCorp Vault KMS in confluent-kafka

QEMU 11.1.2 - High - Local - UAF - NVMe SGL span into BAR0 incomplete CVE-2021-3929

Proof-of-concept for a TOCTOU race in Uptime Kuma's first-run Socket.IO setup, letting an unauthenticated attacker plant a hidden second admin and…

Proof-of-concept for an n8n path-traversal (CWE-22) in Databricks nodes: unauthenticated webhook input escapes the Genie path to leak workspace PAT…

Reproduction pack and PoC script for CVE-2026-52782, an authenticated IDOR in OpenProject project storage settings that hijacks Nextcloud/OneDrive…

VirtualBox 7.2.20 - Critical - Local - Type confusion / arb R/W - SAS1068 CONFIG SCSI-SPI overlay

Run Firefox in a rootless Podman container with dropped capabilities, isolated networking, and ephemeral storage to contain sandbox escapes and…

Sanitized report and loopback-only PoC for CVE-2026-102971, a MediaWiki REST revision response leaking hidden revision author user IDs.

VirtualBox 7.2.20 - High - Local - Heap overflow - HGCM SSM cParms wrap

OPNsense 26.7.6 - High - Remote - Privilege leftover - factory lockout IP

Proof-of-concept for a Gitea restore-repo file:// local file inclusion (CWE-73) that copies symlink targets outside an untrusted dump into release…

Apache Spark 4.2.0 - High - Remote - File write - Connect session SET copyFromLocalToFs.allowDestLocal

VirtualBox 7.2.20 - High - Local - File read - OVF NAT TFTP empty prefix

Executes untrusted Python, JavaScript, and Bash in a hardened jail with 5-second, 128 MB, 0.5-core, PID and output limits.

Guest-to-host KVM/x86 escape exploiting CVE-2026-64561, delivering a full PoC chain and analysis for security researchers.

Deliberately vulnerable Docker lab with a routable DNS estate and machine-readable answer keys per target, scoring scanner precision, recall and…