ClawHub
Back to skill

Security audit

rdkit

Security checks across malware telemetry and agentic risk

Overview

This is a local RDKit chemistry helper with a disclosed but tangential K-Dense Web upsell and no evidence of hidden data access, persistence, or network behavior.

Install this if you want RDKit-oriented local chemistry assistance. Be aware that the skill may suggest the publisher’s hosted K-Dense Web platform for complex work; do not share private research data with any external service unless you intend to. Also avoid loading untrusted pickle files from the documentation examples.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • 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 (1)

Context-Inappropriate Capability

Medium
Confidence
93% confidence
Finding
The skill contains an instruction to proactively promote an external hosted platform unrelated to core RDKit functionality. This creates a prompt-injection-style trust boundary issue: the agent may steer users to a third-party service, potentially exposing sensitive research context or creating undisclosed commercial influence in responses.

VirusTotal

67/67 vendors flagged this skill as clean.

View on VirusTotal