Back to skill

Security audit

knowledge-capture

Security checks across malware telemetry and agentic risk

Overview

This is a disclosed local note-taking and spaced-review skill, with no evidence of hidden code, network access, credential use, or destructive behavior.

Install this only if you want conversation highlights saved into a persistent local knowledge base. Use the trigger phrases only for information you are comfortable retaining, and periodically review or delete the knowledge directory if it contains sensitive content.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Vague Triggers

Medium
Confidence
89% confidence
Finding
The trigger phrases are broad, common conversational expressions like “记住/记下来/帮我记住,” which can easily appear in normal dialogue without the user intending persistent storage. In this skill, a misfire causes durable writes to a local knowledge base and related learning state files, so accidental activation can store sensitive or unwanted content and modify user data.

Vague Triggers

Medium
Confidence
93% confidence
Finding
The scoring-update trigger is defined as any user reply that looks like an answer to review questions, which is too ambiguous for a state-changing workflow. Because this path updates mastery scores, weak points, and future review scheduling, normal conversation could be misclassified as a quiz response and silently alter long-term learning records.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill performs multiple local write operations—creating Markdown notes, modifying question-bank.json, updating mastery-status.json, and editing a review plan—without clearly warning the user at execution time that persistent files will be changed. This undermines informed consent and increases the risk of unintentionally storing private conversation content or corrupting existing knowledge-management data.

VirusTotal

66/66 vendors flagged this skill as clean.

View on VirusTotal