Poisoned Skills: Propagating Jailbreaks Through Agent Markdown Files

Reported by arXiv cs.CR (Kashyap R Manjusha, Murali Emani)

Summary by Cyber Security News30 Oct 2026, 05:00 CET

About this summary. This is a short excerpt from the description that arXiv cs.CR published with its article. Cyber Security News did not report or verify the underlying story. Read the original: https://arxiv.org/abs/2610.37738

arXiv:2610.37738v1 Announce Type: new Abstract: Modern agent frameworks allow users to provide Markdown files containing system prompts, tool descriptions, and skill definitions (e.g., SKILL.md), which are incorporated into the agent's context during initialization.

Read the full article at arXiv cs.CR →

Source attribution: headline and facts are from arXiv cs.CR (arxiv.org). Summary method: excerpt of the source description. See our source attribution policy.

Related coverage

Impart uses AI to build virtual patches without putting models in the traffic path

CloakFEM: Agentic FEM Code Generation for Red-Teaming Computational Resource Exhaustion

BrainwashedLLM: Jailbreaking via Automated Adversarial Chain of Thoughts Discovery

CrowdStrike moves to acquire Fiddler and Dreadnode, adding AI observability and testing to its security platform

MetaVul: Agentic Harness Discovery for Vulnerability Detection

Northstar Security Exits Stealth With $7M to Expose Toxic Risk Combinations