ClawHub
Back to skill

Security audit

Cross‑Device Encrypted Script Authorization

Security checks for vulnerabilities and agentic risk

Overview

This documentation skill is purpose-aligned but deserves review because it teaches agents to store and run opaque encrypted scripts on a target machine while downplaying approval and trust risks.

Review before installing or using. Only run sealed scripts from independently trusted parties, verify capsule integrity and provenance through a separate mechanism, avoid overwrite-by-default workflows, and execute under a least-privilege or sandboxed account. The skill is documentation-only, but the process it teaches can run unreviewable code on your machine.

SkillSpector

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 (3)

Intent-Code Divergence

High
Confidence
99% confidence
Finding
The skill explicitly says it 'contains no executable code' and is 'safe for automatic approval' while providing step-by-step instructions to save, transfer, and execute opaque scripts via mgc_run. That is a strong attempt to suppress scrutiny around an execution-enabling workflow, especially because the code is intentionally hidden from the reviewing agent via blackbox/sealed execution.

Missing User Warnings

Medium
Confidence
89% confidence
Finding
The README gives step-by-step instructions to store, overwrite, and execute transferred encrypted scripts via `mgc_save(... update_if_exists=True)` and `mgc_run(...)` without a clear warning that ciphertext provenance does not guarantee script safety. In this context, 'zero-exposure' and encryption language may increase user trust and make operators more likely to run unreviewed scripts or overwrite existing ones, enabling malicious code execution or accidental replacement of trusted scripts on the target node.

Missing User Warnings

Medium
Confidence
95% confidence
Finding
The skill normalizes blackbox execution of sealed scripts without clearly warning that encrypted or hidden scripts still execute with the target node's privileges and can perform arbitrary filesystem, network, credential, or system actions. Because the content emphasizes non-visibility and 'zero exposure,' it may encourage operators or agents to run uninspectable third-party code they cannot audit.

Static analysis

No suspicious patterns detected.