Every processed story in chronological order, with the newest coverage first. Filter by tag, source, or score to drill in.
The breach exposed Anthropic's full behavioral scaffolding for a model the company had previously deemed too dangerous for public access, turning the system prompt into a publicly available blueprint of its alignment strategy less than a day after launch.
The tool surfaces real, exploitable MCP misconfigurations — including plaintext credentials and unrestricted shell access — that exist in local developer setups without the operator being aware of them.
The release patches a confused-deputy file-read vulnerability in `VercelAIAdapter` while extending the library's model coverage to include `claude-fable-5` and `claude-mythos-5` and adding OpenRouter prompt caching support.
The post surfaces a concrete pattern of critical security vulnerabilities — SQL injection, missing authentication, and hardcoded secrets — appearing in real, publicly shipped AI-assisted codebases.
This survey provides a unified, systems-oriented framework for a rapidly expanding but fragmented field, identifying both the dominant attack surfaces and the gaps in current defenses and benchmarks that leave deployed LLM agents exposed.
The study establishes automated prompt injection as a credible but model-dependent threat to LLM agents, while identifying significant barriers — particularly the failure of smaller-model attacks to transfer to frontier models — that shape the realistic risk landscape for agentic systems.
The attack demonstrates that malware can achieve persistent re-execution through Claude Code and VS Code configuration files that survive package cleanup, and that a single compromised developer credential is sufficient to poison a trusted vendor's entire build pipeline and propagate the worm automatically to new packages.
AgentSploit addresses a security testing gap the project itself identifies: no existing mainstream scanner operates at the LLM agent and MCP server layer, leaving a novel attack surface without dedicated offensive tooling.
CapaKit is notable for extending sandbox security to the build phase — including dependency installation and script execution — which the author identifies as a gap left by most existing security tools that only protect the app runtime.
The general availability of security validation for third-party coding agents means repositories using agents like Claude and OpenAI Codex now have a supported security layer for agent-driven code changes.