Security audit

Afrexai Cybersecurity Engine

Security checks for vulnerabilities and agentic risk

Overview

This security skill is mostly purpose-aligned, but it asks for command execution and broad scanning/workflow authority without enough clear limits or user-control checkpoints.

Install only if you intend to use this for authorized security work. Treat it as a review-required automation aid: approve targets explicitly, review commands before execution, avoid production-impacting scans unless authorized, and do not assume its validation or human-review claims are enforced by the skill.

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: 89% confidence
Finding: The skill claims it performs strict input validation and escaping, but the document exposes broad input-driven capabilities including read/exec and scanning workflows without any concrete validation rules, allowed command set, parameter schema enforcement, or sandbox constraints. In a security-focused skill, this creates a dangerous trust gap: users or orchestrators may assume inputs are safely constrained when they are not, increasing the chance of command injection, unsafe target handling, or misuse of downstream tools.

Intent-Code Divergence

Medium

Confidence: 85% confidence
Finding: The skill states outputs are manually reviewed for safety, but the rest of the file presents it as an automated agent skill whose results can be directly used. This mismatch can cause users to overtrust generated findings, remediation commands, or scan outputs as if a human had validated them, which is especially risky for security operations where false positives, destructive commands, or sensitive disclosures can have real impact.

Vague Triggers

Medium

Confidence: 80% confidence
Finding: The activation language is broad enough to match many generic 'security' requests, which can cause the skill to trigger outside a narrowly intended context. Because the skill advertises powerful capabilities, including command execution and external interactions, overbroad activation increases the risk of unexpected invocation, unnecessary access to local data, or running scans/actions the user did not specifically request.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The skill advertises file handling, external API integration, and command execution, but does not present a prominent up-front warning about possible effects on user data, credentials, network targets, or system state. In context, this is more dangerous because the skill is framed as a cybersecurity engine, making invasive actions seem normal; users may not realize that scans, commands, or file operations can alter systems, expose sensitive data, or hit unauthorized targets.

Static analysis

No suspicious patterns detected.