ClawHub
Back to skill

Security audit

Paper Code Joint Analysis

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent paper-and-code analysis workflow that writes local analysis artifacts and an optional static reader, with no artifact-backed evidence of hidden exfiltration or destructive behavior.

Install only if you are comfortable letting Codex read the paper and target repository, create analysis files, and run the bundled validation/reader scripts. Keep runs in a dedicated project directory, and only allow dependency installation, experiment execution, or KaTeX font downloads when you explicitly want those actions.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • MCP Least PrivilegeUnderdeclared Capability, Wildcard Permission, Missing Permission Declaration
  • 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)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill instructs the agent to read and write files, invoke local validation/build scripts, and use network access to download KaTeX fonts, but the metadata does not declare any permissions. This creates a capability/expectation mismatch that can lead to over-privileged execution, weak policy enforcement, or reviewers/users being unaware that the skill may mutate local artifacts and access the network.

VirusTotal

61/61 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.obfuscated_code

Potential obfuscated payload detected.

Warn
Code
suspicious.obfuscated_code
Location
assets/reader-template/vendor/mermaid.min.js:371