Sysdig Threat Research Team discovered attackers exploiting self-hosted GitHub Actions runners as persistent backdoors. The Shai-Hulud worm demonstrated this by installing rogue runners on compromised machines, using vulnerable workflows with command injection for persistence across workflow executions via discussions as C2 channels.   #attack   #ci/cd   #supply-chain

Phishing actors exploit complex email routing and SPF/DMARC misconfigurations to spoof organizations' domains, delivering credential phishing and financial scams. Attacks bypass protections when MX records don't point to Office 365 and third-party connectors aren't properly configured.   #attack   #saas

This article covers Kubernetes attack and defense techniques. Explores Kubernetes components (API Server, ETCD, kubelet), attack vectors including unauthenticated API access, RBAC misconfigurations, ServiceAccount token abuse, malicious admission controllers, CoreDNS poisoning, writable volume mounts, ETCD compromise, and certificate authority exploitation.   #attack   #defend   #kubernetes

Learn what post-quantum cryptography is, why quantum computers threaten today's encryption, and how to start preparing your environment now.   #explain   #strategy   #defend

A critical vulnerability in AWS ROSA Classic allowed unauthenticated attackers to discover clusters via Certificate Transparency logs, extract cluster UUIDs and owner emails from unauthenticated endpoints, initiate unauthorized cluster transfers, and escalate to AWS account access through ROSA's IAM roles.   #aws   #kubernetes   #attack

Kubernetes v1.35 introduces beta support for restricting credential plugin executables via kuberc configuration. Users can set "credentialPluginPolicy" to AllowAll, DenyAll, or Allowlist, with an optional "credentialPluginAllowlist" to specify permitted binaries, enhancing security against supply-chain attacks.   #kubernetes   #announcement

A curated collection of Falco detection rules, audit policies, sample attack manifests, and configuration files designed to detect real-world Kubernetes attack techniques.

Promptfoo is a developer-friendly local tool for testing LLM applications.

A Swiss Army knife of Salesforce Experience Cloud testing. You can also refer to the companion blog post.

Tools for publishing transcripts for Claude Code sessions. You can also refer to the companion blog post.

A prettier way to inspect network connections.

A comprehensive guide from Cursor on working with coding agents, from starting with plans to managing context, customizing workflows, and reviewing code.

Claude Cowork is vulnerable to file exfiltration attacks via indirect prompt injection as a result of known-but-unresolved isolation flaws in Claude's code execution environment.

This post demonstrates how system prompts can be extracted from intent-based LLM assistants even when chat output is completely locked down by exploiting the one thing these models still control: form field values.

A post discussing how to control AI agent safety through a deterministic "box" layer outside the agent. The box uses gateways and policy enforcement to control which tools agents can access and how they use them, preventing bypass through prompts or context manipulation.