Security audit

科普智创审校助手

Security checks across malware telemetry and agentic risk

Overview

This science-writing helper appears legitimate, but it should be reviewed because it can create and share Feishu documents and save document links locally without clear final approval.

Install only if you want the agent to use web search plus Feishu tooling. Before running the full workflow, confirm the Feishu workspace, parent node, account, sharing permissions, content format, and whether memory/YYYY-MM-DD.md should be written; use preview-only mode if you do not want remote document creation or retained links.

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)

Description-Behavior Mismatch

Medium

Confidence: 95% confidence
Finding: The skill's declared purpose is content collection, drafting, and review, but the workflow also performs external side effects by creating Feishu wiki nodes and persisting links to local memory. This expands the skill's capabilities beyond user-visible expectations and can cause unauthorized document creation or unintended data persistence if auto-invoked.

Intent-Code Divergence

High

Confidence: 97% confidence
Finding: The file contains contradictory instructions: one section requires XML and forbids Markdown for `feishu_doc`, while a later section says the content should be pure Markdown. In agent settings, conflicting tool-use instructions are dangerous because they can cause malformed writes, bypass intended validation/escaping, or lead the agent to choose an unsafe interpretation when sending content to an external document system.

Vague Triggers

Medium

Confidence: 83% confidence
Finding: The activation text includes broad language such as 'or similar intent,' which can cause the skill to trigger for loosely related user requests. Over-broad activation increases the chance that document creation, web collection, or persistence behaviors run in contexts the user did not specifically request.

Missing User Warnings

Medium

Confidence: 96% confidence
Finding: The skill instructs the agent to write document links into a local `memory/YYYY-MM-DD.md` file without any user-facing notice or consent. Silent local persistence can store user-associated activity across sessions and violates the principle of minimizing side effects in content-generation skills.

Ssd 3

Medium

Confidence: 94% confidence
Finding: Persistently recording generated document links in memory files can create a durable log of user-related publishing activity, potentially exposing topics, dates, and document locations. In this skill context, the risk is increased because the links are tied to generated content and external Feishu resources, making later disclosure or misuse more plausible.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.