Six Dim Evaluator

• Outcome: Review
• Latest audit: May 29, 2026, 9:32 PM
• Version:
  0.1.0

Overview

This appears to be a real evaluator skill, but it asks for broad execution/data-handling authority while its scoring and retention behavior are under-scoped.

Install only if you are comfortable granting evaluator-style access to read skill projects and potentially run tests. Treat scores as approximate until placeholder scoring is fixed, and require clear opt-in controls before any API calls, log analysis, or database retention are enabled.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (4)

Description-Behavior Mismatch

Medium

• Confidence: 96% confidence
• Finding: The skill advertises automated six-dimension evaluation and report generation, but the implementation largely returns heuristic scores derived from a few file existence checks plus fixed placeholder values. This creates a trust/integrity risk: downstream users may rely on the output as an authoritative assessment when it is not actually measuring most claimed dimensions, leading to incorrect decisions about skill quality or safety.

Intent-Code Divergence

Medium

• Confidence: 97% confidence
• Finding: The dimension functions are presented as meaningful evaluations, yet most sub-scores are fixed constants such as 0.70 or 0.75 rather than results of real analysis. In an evaluation engine, this is dangerous because it produces misleadingly precise scores that can conceal weak skills, inflate confidence, and undermine any governance or ranking process built on these outputs.

Missing User Warnings

Medium

• Confidence: 90% confidence
• Finding: The skill explicitly states that evaluation data will be stored in an evaluation database for trend analysis, but it does not describe what data is retained, how long it is kept, who can access it, or whether consent is obtained. In a skill that evaluates other skills and may process logs, reports, and metadata, silent retention can expose potentially sensitive operational or user-related information and creates avoidable privacy and compliance risk.

Missing User Warnings

Medium

• Confidence: 93% confidence
• Finding: The documented workflow includes querying an external API and analyzing usage logs, but it provides no warning that data may be transmitted to third parties or that internal logs may be processed. Because this skill has Bash and Exec available and is framed as an automated evaluator, operators could run it on real projects without realizing that potentially sensitive metadata, usage patterns, or identifiers may leave the local environment.

VirusTotal

66/66 vendors flagged this skill as clean.