Claw Shell 1.0.0

• Outcome: Review
• Latest audit: May 29, 2026 at 9:22 PM UTC
• Version:
  1.0.0

Overview

This skill openly provides local shell access, but its guardrails are weak and its tmux-only safety boundary is not reliable enough for automatic installation.

Review this before installing as it effectively gives an agent local shell access. Only use it in a sandbox or trusted workspace, assume allowed commands can affect local files and data, and do not rely on its current denylist as a complete safety control.

SkillSpector

By NVIDIA

Vulnerability Patterns

• Data Exfiltration: External Transmission, Env Variable Harvesting, File System Enumeration
• Prompt Injection: Instruction Override, Hidden Instructions, Exfiltration Commands
• Privilege Escalation: Excessive Permissions, Sudo/Root Execution, Credential Access
• Supply Chain: Unpinned Dependencies, External Script Fetching, Obfuscated Code
• Excessive Agency: Unrestricted Tool Access, Autonomous Decision Making, Scope Creep

Findings (1)

Missing User Warnings

Medium

• Confidence: 96% confidence
• Finding: The skill executes arbitrary shell commands provided via input after only a weak blocklist check, with no explicit user confirmation or meaningful restriction on allowed operations. In the context of an agent skill, this is dangerous because an LLM, prompt injection, or a mis-specified task could cause execution of destructive, data-exfiltrating, or persistence-establishing commands that do not match the small denylist.

VirusTotal

64/64 vendors flagged this skill as clean.