Attention-OS

Security checks across malware telemetry and agentic risk

Overview

This is a coaching-style prompt skill that changes response style but does not add code, data access, persistence, or privileged actions.

Install this if you want the agent to use a reflective coaching mode. Expect it may make ordinary advice or planning conversations less direct, but the reviewed artifact does not ask for sensitive access or perform actions beyond normal skill instructions.

SkillSpector

By NVIDIA

Vulnerability Patterns

Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (2)

Vague Triggers

Medium

Confidence: 88% confidence
Finding: The skill description is broadly phrased and lacks clear activation boundaries, which can cause the agent to apply this behavioral overlay in situations beyond the author's apparent intent. In a multi-skill or safety-sensitive environment, ambiguous scope can lead to response hijacking, suppression of more appropriate system behaviors, or unintended interference with higher-priority instructions.

Vague Triggers

Medium

Confidence: 93% confidence
Finding: The instruction applies to virtually any user thought, idea, plan, or advice request, giving the skill near-global behavioral scope. This makes it more dangerous because it can systematically redirect normal assistant behavior away from direct answers and toward the skill's preferred interaction pattern, potentially degrading reliability and creating prompt-scope conflicts across unrelated tasks.

VirusTotal

63/63 vendors flagged this skill as clean.

View on VirusTotal