Security audit

DeepTutor 轻量辅导台

Security checks across malware telemetry and agentic risk

Overview

This tutoring skill is coherent and purpose-aligned, but users should understand it keeps local student learning profiles over time.

Install only if you want a tutor that remembers student weaknesses and progress in local files. Use non-sensitive student IDs, avoid storing unnecessary personal details, check who can access the workspace, and periodically review or delete tutor_memory files if the profile should not be retained.

SkillSpector

By NVIDIA

Vulnerability Patterns

Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code

Findings (7)

Vague Triggers

Medium

Confidence: 91% confidence
Finding: The trigger list is unusually broad and includes common tutoring phrases such as personalized tutoring, mastery tracking, and quote-based explanation, which can cause the skill to activate in contexts where the user did not intend to invoke persistent profiling or specialized tutoring behavior. In this skill, accidental activation is more concerning because the workflow includes memory writes and routing behavior, so overbroad matching can lead to unintended data retention and scope expansion.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The README explicitly states that student memory is persisted in files under tutor_memory/ and carried across workspaces, but it does not present a clear user-facing warning, consent mechanism, retention policy, or deletion process. Because the stored content is a student profile with weaknesses, mastery probabilities, and learning history, accidental or uninformed collection can expose sensitive educational and behavioral data over time.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill description emphasizes personalization and long-term tutoring but does not clearly warn users that it persistently stores student learning history, weakness profiles, and mastery data across sessions. Because this includes user-specific educational profiling, the omission creates a transparency and consent problem and increases the risk of collecting more sensitive behavioral data than users realize.

Missing User Warnings

Low

Confidence: 88% confidence
Finding: The skill instructs the agent to generate local `.html` files or use persistent file outputs for visualization without clearly warning about local file creation in the user-facing description. While lower severity than profiling, undisclosed file generation can surprise users, leave residual artifacts on disk, and unintentionally expose educational content or personal context to others with workspace access.

Missing User Warnings

Medium

Confidence: 94% confidence
Finding: The skill explicitly directs retrieval from both personal and subscribed knowledge bases, but the description does not clearly disclose that broader knowledge repositories may be searched during operation. This can expose private or licensed material unexpectedly and may cause users to share prompts under the false assumption that only their directly provided content or personal library is being accessed.

Ssd 3

Medium

Confidence: 97% confidence
Finding: The skill is designed to persistently collect and reuse student-specific learning history, weaknesses, preferences, and mastery scores by default. In a tutoring context this is functionally relevant, but it still constitutes sustained profiling across sessions; if misused or insufficiently disclosed, it can reveal sensitive educational performance patterns and enable unwanted tracking.

Ssd 3

Medium

Confidence: 98% confidence
Finding: These instructions repeatedly require reading back, updating, and writing detailed student profile and performance information, including mastery probabilities, error types, evidence trails, and longitudinal learning preferences. Even though the intent is educational personalization, this level of persistent behavioral profiling increases privacy risk, especially if files are accessible to others in the workspace or retained indefinitely.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.