Security audit

model-troubleshooter

Security checks for vulnerabilities and agentic risk

Overview

This troubleshooting skill appears useful, but it needs Review because it can read sensitive local configuration and logs and may change or restart gateway/model settings without clear user approval.

Install only if you are comfortable letting the skill inspect model/gateway configuration and logs. Before using it, ask the agent to show a dry run, redact secrets, and get explicit approval before changing configs or restarting services.

SkillSpector

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access

Findings (4)

Intent-Code Divergence

Medium

Confidence: 90% confidence
Finding: The skill claims secrets will never be exposed, yet it instructs the agent to inspect full configuration and determine API key status. That creates a mismatch between stated safety guarantees and actual behavior, increasing the risk that sensitive credential material is accessed, processed, or surfaced in logs or outputs during troubleshooting.

Vague Triggers

High

Confidence: 98% confidence
Finding: The trigger condition 'Any model-related complaint' is so broad that the skill can activate on vague user language and begin sensitive diagnostics automatically. In this skill, activation leads to config reads, log inspection, and potentially destructive actions, so ambiguous triggering materially raises the chance of unintended execution.

Missing User Warnings

High

Confidence: 97% confidence
Finding: The skill states that no manual intervention is needed and later authorizes config patches and gateway restarts, but it does not provide a clear upfront warning that system state may be modified. Automatic operational changes can disrupt service, break configurations, or switch providers without informed user consent.

Missing User Warnings

Medium

Confidence: 95% confidence
Finding: The skill directs the agent to read complete configuration, inspect logs, and assess API key status without clearly warning the user that sensitive local data will be accessed. This creates a privacy and secret-handling risk because users may not realize the extent of data collection involved in the automated scan.

Static analysis

No suspicious patterns detected.