Security audit

Code Quality Analyzer (FightingDao)

Security checks across malware telemetry and agentic risk

Overview

This appears to be an internal code-quality reporting skill, but it can overwrite reporting data, delegate code review to another model, and reference notification channels without enough user control or disclosure.

Install only in the intended internal environment after reviewing the referenced local scripts. Before running it, confirm the exact period, team, projects, database target, backup/rollback plan, notification recipients, and whether secondary-model review is allowed for the repository content.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (5)

Description-Behavior Mismatch

Medium

Confidence: 87% confidence
Finding: The skill manifest describes analysis and database synchronization, but the body also introduces Teams and email notification capabilities. This scope expansion can cause the agent to perform outbound communications the user did not explicitly request, increasing the risk of unintended data disclosure or surprise side effects.

Description-Behavior Mismatch

Medium

Confidence: 84% confidence
Finding: The skill is presented as a code-quality analysis and sync tool, but later instructions include modifying frontend/backend applications and implementing pages/modules. That broadens the authority of the skill from analysis into codebase modification, which can trigger unintended changes to production systems beyond user expectations.

Intent-Code Divergence

High

Confidence: 96% confidence
Finding: The document claims code review does not call an external AI API, yet elsewhere instructs spawning a different model session and passing code changes for review. This inconsistency can mislead users about where sensitive code is sent and creates a hidden data-transfer path to another model or service.

Missing User Warnings

Medium

Confidence: 89% confidence
Finding: The skill includes deletion and overwrite operations against database records for a given period, but does not clearly require user confirmation or present strong risk warnings before performing destructive actions. In an agent setting, this can lead to unintended data loss or corruption if the period, scope, or environment is wrong.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The skill describes automatically switching models and sending code diffs for review without clearly informing the user that repository content will be transmitted to another model/session. This creates a data exfiltration and confidentiality risk, especially for proprietary source code or embedded secrets in diffs.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.