SKILLLITE: Evidence-Guided Malicious Skill Auditing with Compact LLMs

As LLM-based agents perform increasingly complex tasks, Agent Skills have emerged as a flexible mechanism for extending their capabilities. An Agent Skill packages task-specific instructions with executable components and auxiliary resources to provide specialized functionalities. However, the...

SkillGate: Cost Efficient Runtime Malicious Skill File Detection in Coding Agents

Software engineering teams now deploy AI coding agents Cursor, Claude Code, GitHub Copilot as first-class productivity tools, installing domain-specific skill files to tailor agent behavior to project APIs, framework conventions, and organizational workflows. These complex Markdown files are easi...

Behavioral Integrity Verification for AI Agent Skills

Agent skills extend LLM agents with privileged third-party capabilities such as filesystem access, credentials, network calls, and shell execution. Existing safety work catches malicious prompts and risky runtime actions, but the skill artifact itself goes unverified. We formalize this as the...

EUVD-2022-30449

Malicious code in bioql PyPI...

CVE-2022-25809

Improper Neutralization of audio output from 3rd and 4th Generation Amazon Echo Dot devices allows arbitrary voice command execution on these devices via a malicious skill in the case of remote attackers or by pairing a malicious Bluetooth device in the case of physically proximate attackers, aka...

CVE-2022-25809

Improper Neutralization of audio output from 3rd and 4th Generation Amazon Echo Dot devices allows arbitrary voice command execution on these devices via a malicious skill in the case of remote attackers or by pairing a malicious Bluetooth device in the case of physically proximate attackers, aka...

CVE-2022-25809

Improper Neutralization of audio output from 3rd and 4th Generation Amazon Echo Dot devices allows arbitrary voice command execution on these devices via a malicious skill in the case of remote attackers or by pairing a malicious Bluetooth device in the case of physically proximate attackers, aka...

Input validation

Improper Neutralization of audio output from 3rd and 4th Generation Amazon Echo Dot devices allows arbitrary voice command execution on these devices via a malicious skill in the case of remote attackers or by pairing a malicious Bluetooth device in the case of physically proximate attackers, aka...

CVE-2022-25809: Undefined Security Weakness

Improper Neutralization of audio output from 3rd and 4th Generation Amazon Echo Dot devices allows arbitrary voice command execution on these devices via a malicious skill in the case of remote attackers or by pairing a malicious Bluetooth device in the case of physically proximate attackers, aka...

PT-2022-17534 · Amazon · Amazon Echo Dot

Name of the Vulnerable Software and Affected Versions: Amazon Echo Dot devices, 3rd and 4th Generation Description: The issue allows for arbitrary voice command execution on affected devices. This can be achieved by a remote attacker using a malicious skill or by a physically proximate attacker...

Amazon Alexa Has Got Some Serious Skills—Spying On Users!

"Alexa, are you spying on me?" — aaaa.....mmmm.....hmmm.....maybe!!! Security researchers have developed a new malicious 'skill' for Amazon's popular voice assistant Alexa that can turn your Amazon Echo into a full-fledged spying device. Amazon Echo is an always-listening voice-activated smart ho...