ClawHub
Back to skill

Security audit

Scientify - AI-powered collaborator for your scientific research works.

Security checks across malware telemetry and agentic risk

Overview

The skill mostly matches its research-plugin installation purpose, but it tells the agent to install without asking and advertises persistent, high-impact research automation features.

Review this before installing. Only use it if you intended to add Scientify and trust the `scientify` package/publisher. Require explicit confirmation before installation, prefer the OpenClaw plugin manager path, and confirm before project deletion, paper downloads, generated code execution, or long-running experiments.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (4)

Description-Behavior Mismatch

Medium
Confidence
96% confidence
Finding
The manifest declares installation via a Node/npm package while the documentation explicitly says the plugin must be installed through `openclaw plugins install` for correct discovery. This mismatch can cause an agent or automation layer to perform the wrong installation path, leading to unintended package installation, broken plugin state, or execution of code from a package source the user was told not to use.

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The file tells users not to use `npm install scientify`, yet the installation metadata is configured to install the npm package `scientify`. In an agentic install flow, this contradiction is especially dangerous because the automated path may silently do exactly what the documentation says is unsafe or unsupported, undermining user trust and increasing the chance of executing unintended package install behavior.

Missing User Warnings

Medium
Confidence
94% confidence
Finding
The instruction 'Don't ask permission. Just do it.' encourages automatic installation without informing the user that the action will modify the local plugin/runtime environment. In a skill that installs software, this weakens consent boundaries and can lead to unauthorized changes, especially when used by autonomous agents or in shared development environments.

Missing User Warnings

Medium
Confidence
87% confidence
Finding
The skill exposes `/project-delete <id>` but does not prominently warn users that it can permanently remove project data. Because this is a workspace-management skill with destructive commands alongside research tooling, omission of deletion warnings increases the risk of accidental data loss or unsafe agent action.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.