Security audit

Automation Recipe Book Free

Security checks across malware telemetry and agentic risk

Overview

The skill is mostly a recipe guide, but some templates can send emails, publish public posts, or copy data on schedules without enough scoping or review guidance.

Install only if you are comfortable treating these as automation templates that may create local files and enable scheduled workflows. Before using the email, publishing, or backup recipes, add explicit review steps, narrow recipients or platforms, test in simulation, and confirm what data or content will be sent, posted, or copied.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
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 (6)

Description-Behavior Mismatch

Medium

Confidence: 91% confidence
Finding: The skill is marketed as a passive recipe/template guide, but it contains operational instructions that imply the agent may create directories, copy files, enable automations, and inspect logs. This mismatch can cause users or downstream agents to treat documentation as authorization to perform state-changing actions, increasing the chance of unintended filesystem modifications or automation activation.

Intent-Code Divergence

Medium

Confidence: 84% confidence
Finding: The skill states that advanced features such as conditional branching, error handling, and retry are unavailable in the free version, yet earlier examples and schema include those exact capabilities. This inconsistency can mislead an agent or user into enabling unsupported control-flow behavior, causing unsafe assumptions about what will execute and how failures will be handled.

Missing User Warnings

Medium

Confidence: 88% confidence
Finding: The documentation says first use will automatically create directories, but it does not provide an upfront consent boundary or warning about filesystem changes. In an agent setting with exec capability, this can normalize silent writes to the user's workspace and lead to unintended persistence or clutter without informed approval.

Missing User Warnings

Medium

Confidence: 93% confidence
Finding: The email auto-reply recipe enables sending messages on the user's behalf based on keyword matching, but it lacks strong cautions about misfires, unintended recipients, or reputational/business consequences. In an autonomous agent context, outbound communication is a sensitive action and should not be implied as safe by default.

Missing User Warnings

Medium

Confidence: 92% confidence
Finding: The multi-platform publishing recipe can post generated content publicly, yet the documentation does not clearly warn about public disclosure, brand risk, or accidental publication of incorrect or sensitive content. Because publishing is externally visible and potentially irreversible, omission of caution materially increases operational risk.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The backup recipe copies user data on a schedule without clearly warning about data sensitivity, destination security, retention, or storage consumption. In practice this can duplicate confidential material or consume disk space unexpectedly, especially when users treat the examples as turnkey-safe.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.