ClawHub

ScienceClaw: Watch (Live Collaboration)

Security checks across malware telemetry and agentic risk

Overview

This skill is a disclosed scientific research workflow that runs a local ScienceClaw tool, saves generated results, and may include workspace project context in the research topic.

Use this only if you trust the local ScienceClaw installation it will execute and are comfortable using your Anthropic API key. Review memory.md before running if it may contain sensitive project context, and expect generated summaries and figures to remain on disk under run_exports.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
Findings (2)

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The skill explicitly states that results and figures are saved to disk, but it does not clearly warn the user that execution may persist potentially sensitive inputs or derived outputs in the workspace. In addition, later instructions tell the operator to read workspace memory, so the overall workflow can access and retain contextual data without an explicit user consent step or disclosure.

Ssd 3

Medium
Confidence
96% confidence
Finding
The skill instructs the agent to read memory.md, append any discovered project context directly into the topic string, and then summarize the session back to chat. That creates a straightforward natural-language exfiltration path where stored workspace context can be disclosed to the user or downstream tools even if it was not part of the user's current request.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal