Prompt Optimizer

Security checks across malware telemetry and agentic risk

Overview

This skill is a prompt-writing and profile-polishing helper with disclosed user-directed file reading and no executable payload or hidden persistence.

Safe to install for drafting AI roundtable personas and polishing owner profiles. When using polish with --file, choose only files you are comfortable placing in the agent context, and review generated profiles before using them publicly.

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: 83% confidence
Finding: The `polish --file=...` workflow encourages reading arbitrary local files containing owner profiles without any warning about sensitive data exposure or scope restriction. In practice, users may point the skill at files holding personal, corporate, or strategic information, and the skill's broad Read/Glob/Grep permissions increase the chance of over-collection or unintended disclosure into model context.

VirusTotal

56/56 vendors flagged this skill as clean.

View on VirusTotal