SKILL审查

• Outcome: Pass
• Latest audit: May 29, 2026 at 8:36 PM UTC
• Version:
  1.0.0

Overview

This is a local skill-folder review tool with a disclosed, limited auto-fix mode that creates standard subdirectories but does not show hidden, destructive, credential, network, or persistence behavior.

Install this only if you want an agent helper for reviewing skill directories. Use the read-only analyze command first, and run the review-loop mode only on a skill folder you are comfortable letting it modify by creating missing standard subdirectories.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Trigger Abuse: Overly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
• MCP Tool Poisoning: Hidden Instructions, Unicode Deception, Parameter Description Injection
• 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

Findings (2)

Description-Behavior Mismatch

Medium

• Confidence: 96% confidence
• Finding: The skill is presented as a static reviewer, but the '自我迭代模式' instructs running a review-repair loop that automatically fixes issues until none remain. This is a capability expansion from analysis into autonomous modification, which can lead to unauthorized file changes, destructive edits, or repeated mutation of a target skill without clear human approval.

Vague Triggers

Medium

• Confidence: 87% confidence
• Finding: The trigger phrases are broad enough to match common user requests such as improving or reviewing a skill, which can cause this skill to activate outside its intended scope. In an agent system, over-broad routing can expose internal analysis or modification behavior unexpectedly and interfere with other skills, creating prompt-scope confusion and unsafe task execution paths.

VirusTotal

62/62 vendors flagged this skill as clean.