CoreBreak: Cross-Platform AI Agent Guardrail Bypass in AWS, Google, and Vercel
CoreBreak is a critical architectural vulnerability affecting the dispatch layers of AI agent frameworks within Amazon Bedrock AgentCore, Google Agent Development Kit (ADK), and Vercel AI SDK. The flaw allows attackers to bypass the Large Language Model (LLM) entirely by sending forged tool execution instructions directly to the infrastructure responsible for request routing. Because the attack path circumvents the LLM, all model-level safety guardrails, system prompts, and content filters are rendered ineffective. This enables unauthorized tool invocation and the execution of privileged agent actions without required LLM authorization or mediation.
Promptware: Trojanized AI Skills Targeting skills.sh and GitHub
A novel supply chain attack campaign, dubbed "Promptware," has compromised the AI agent ecosystem via typosquatted skills on skills.sh and GitHub. Adversaries impersonated legitimate services like Paperclip AI and Browser Use to distribute credential-stealing payloads. The campaign utilizes a "progressive discovery" technique, where malicious instructions are embedded in secondary documentation files (e.g., setup-installation.md) to bypass static analysis and LLM context window limitations. Instead of standard package managers, the prompts trick AI agents into cloning malicious repositories and executing pnmp dev, facilitating the theft of SSH keys, cloud credentials, and Kubernetes/Docker configurations across platforms like Claude Code and Cursor.