ClawHub
Back to skill

Security audit

Untrusted Binary Sandbox

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed safety workflow for analyzing untrusted wallet or trading software in isolated sandboxes, with safeguards against using real secrets or production systems.

Install this only if you intend to analyze untrusted software. Review generated Docker files before use, run samples only in a disposable VM or tightly restricted container, and never mount real wallets, SSH keys, browser profiles, cloud credentials, production .env files, or funded accounts.

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
90% confidence
Finding
The default prompt is broadly worded and can trigger the skill for any request to 'plan a staged sandbox observation' of a wallet or trading bot release without requiring explicit safety preconditions, user consent boundaries, or environment constraints. In a high-risk security-analysis context involving untrusted binaries, overly broad invocation text can cause the agent to engage this powerful workflow too readily, increasing the chance of unsafe guidance or misuse on sensitive artifacts.

Static analysis

No suspicious patterns detected.