CYBERSECURITYTRACKER
TRACKING6,506 stories in this site build1,309 vulnerability news stories in this site build
Permanent story citation

Malicious AI agent skills can slip past the scanners built to stop them

This page keeps the story as Cybersecurity Tracker first published it. If the tracker later corrects it, the correction appears below the original and never replaces it.

Back to newsStory 2216

As cited

Copy frozen at (site build).

ai security

Malicious AI agent skills can slip past the scanners built to stop them

AI agent skills, small bundles of code and instructions used by AI coding tools, are being distributed through public marketplaces similar to traditional package repositories. Security scanners designed to detect malicious skills have been found to miss threats, and a marketplace has accumulated over 40,000 listed skills within months of the format's emergence.

Why it matters: Developers integrating third-party AI agent skills into their systems face supply chain risk if existing security controls fail to identify malicious or compromised skills before deployment.

Source published
First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

Malicious AI agent skills can slip past the scanners built to stop them

No summary had been written when this copy was frozen.

First seen by Cybersecurity Tracker

Source attribution

Correction

Correction recorded as of .

ai security

Malicious AI agent skills can slip past the scanners built to stop them

Developers use public marketplaces to add agent skills, small bundles of instructions and scripts, to artificial intelligence (AI) coding tools like Claude Code and OpenAI Codex. Malicious skills in these repositories can evade detection systems that are meant to block harmful code before it executes.

Why it matters: Development teams and platform operators should audit third-party AI agent skills before deployment, as current scanning defenses may not catch malicious code disguised in plain-English instructions or obfuscated scripts.

Source published
First seen by Cybersecurity Tracker

Source attribution

Glossary