Security audit

Lifestyle Assistant Tool Free

Security checks across malware telemetry and agentic risk

Overview

The skill is a coherent personal assistant, but it asks to handle and persist sensitive personal tasks, email summaries, preferences, and history with broad proactive behavior and unclear privacy boundaries.

Install only if you are comfortable with a local assistant storing tasks, notes, email summaries, reminders, preferences, and history under ~/.assistant. Confirm the timezone and working hours before using scheduling, and avoid giving it sensitive email or calendar content unless you understand whether your agent's LLM processing is local or cloud-based.

SkillSpector

By NVIDIA

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 (5)

Intent-Code Divergence

Medium

Confidence: 93% confidence
Finding: The skill claims that all data is stored locally and not uploaded to the cloud, yet it also declares dependence on an LLM API or built-in inference service for core assistant functions such as summarization. This creates a misleading privacy guarantee: users may disclose sensitive emails, tasks, and schedule data believing it remains local when it may be sent to an external model provider.

Vague Triggers

Medium

Confidence: 86% confidence
Finding: The capability coverage text defines an extremely broad activation scope across many generic personal productivity scenarios, without clear boundaries or consent gates. In an agent environment with read/write/exec tools, overly broad triggering can cause the skill to engage on unrelated user content and handle more personal data than the user reasonably expects.

Missing User Warnings

Medium

Confidence: 90% confidence
Finding: The skill describes proactive capture, reminders, and information archiving, but does not present a prominent notice about privacy, retention, or the user impact of continuous collection of personal tasks, emails, preferences, and history. This is risky because the skill also exposes read/write/exec capabilities, increasing the chance of silent collection and persistence of sensitive personal information.

Natural-Language Policy Violations

Medium

Confidence: 78% confidence
Finding: The sample configuration hard-codes a specific timezone as the default without making clear that the user should choose or confirm their own locale. For a scheduling assistant, incorrect defaults can lead to inaccurate reminders, missed meetings, or unintended disclosure through wrongly timed communications.

Natural-Language Policy Violations

Low

Confidence: 74% confidence
Finding: The environment variable example again suggests a fixed timezone value without clarifying that it is only illustrative. In context, this is less severe than data exfiltration issues, but it can still cause user-harm through scheduling mistakes in a tool meant to automate reminders and coordination.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.