Security audit

senior-security

Security checks across malware telemetry and agentic risk

Overview

This skill is not malware, but its security tools appear to be placeholders that can report success without doing real security analysis.

Treat this as placeholder reference material, not a real security audit or pentest toolkit. Do not rely on its clean or zero-finding output for compliance, release decisions, or vulnerability review, and avoid running pentest-style commands except on systems where you have explicit authorization and scope.

SkillSpector

By NVIDIA

Vulnerability Patterns

MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
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 (6)

Description-Behavior Mismatch

Medium

Confidence: 94% confidence
Finding: The script is presented as a comprehensive security auditing tool, but it only checks that a path exists and then returns a hardcoded success result with zero findings. This can create a dangerous false sense of security: users may rely on the tool's clean output and skip real security review, leaving vulnerabilities undetected.

Intent-Code Divergence

Medium

Confidence: 92% confidence
Finding: The top-level documentation and class description claim this is a security auditor for senior security tasks, but the implementation performs no substantive audit. In a security-focused skill, misleading documentation is especially risky because operators may trust the branding and output as evidence of assessment coverage.

Description-Behavior Mismatch

Medium

Confidence: 93% confidence
Finding: The script presents itself as a threat-modeling and security-analysis tool, but it performs only basic path existence validation and then always reports success with empty findings. In a security skill context, this can mislead users into believing an assessment was performed, causing missed vulnerabilities and unsafe decision-making based on false assurance.

Intent-Code Divergence

Medium

Confidence: 95% confidence
Finding: The docstrings, CLI description, and report output all imply meaningful analysis and report generation, yet the code contains no actual threat-modeling logic and unconditionally returns a successful result. Because this skill is marketed for senior security engineering tasks, the mismatch is more dangerous: operators may trust placeholder output during audits, reviews, or design assessments and overlook real risks.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill promotes a 'Pentest Automator' with expert-level automation but provides no warning about authorization, scope control, rate limiting, or possible service disruption. In a security-focused skill, that omission is more dangerous because users may assume offensive testing is safe to run by default, increasing the risk of unauthorized scanning or harmful activity against production systems.

Missing User Warnings

Medium

Confidence: 86% confidence
Finding: Advertising 'automated fixes' without stating that files may be changed can lead users to run the tool on important repositories without understanding the write impact. In a skill that presents itself as a trusted security toolkit, this can cause unintended code changes, broken builds, or risky mass modifications that are hard to review.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal