skill-usefulness-audit

• Outcome: Pass
• Latest audit: Jun 5, 2026 at 11:02 AM UTC
• Version:
  0.3.0

Overview

This is a local skill-auditing tool that reads installed-skill folders and optional evidence files to produce reports, with no evidence of hidden network access, credential theft, or automatic deletion.

Install this only if you want a local audit of installed agent skills. Prefer passing explicit skill roots and sanitized evidence files, especially for history exports, and treat any delete or quarantine recommendation as a manual review prompt rather than an instruction to remove skills automatically.

SkillSpector

By NVIDIA

Vulnerability Patterns

• MCP Least Privilege: Underdeclared Capability, Wildcard Permission, Missing Permission Declaration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (1)

Lp3

Medium

• Category: MCP Least Privilege
• Confidence: 90% confidence
• Finding: The skill advertises `disable-model-invocation: true` but its instructions clearly direct execution of a local Python script that reads and writes multiple files, may inspect usage/history data, and can process optional community inputs. The metadata declares required binaries but does not declare explicit permissions for file, shell, environment, or possible network access, creating a transparency and least-privilege gap that can lead hosts to under-enforce or misjudge the skill's capabilities.

VirusTotal

65/65 vendors flagged this skill as clean.