BanditDB

Security checks across malware telemetry and agentic risk

Overview

BanditDB is a coherent local decision-learning skill, with expected privacy and persistence considerations but no hidden or unsafe behavior in the artifacts.

Before installing, verify the external BanditDB release, Docker image, or SDK source and prefer pinned versions. Use minimal, non-sensitive context vectors, get user consent where appropriate, define retention/deletion practices, and avoid using learned rewards to manipulate users or make high-risk decisions.

SkillSpector

By NVIDIA

Vulnerability Patterns

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
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (1)

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill explicitly promotes behavior-learning and user-targeting use cases such as notification timing, response style, heartbeat frequency, and model/tool routing based on contextual data, but provides no privacy, consent, retention, or minimization guidance. This creates a real risk that integrators will collect and optimize on behavioral or preference data in ways that enable profiling, manipulative nudging, or non-compliant personal data processing.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal