ClawHub
Back to skill

Security audit

Unikraft Cloud Sandbox

Security checks for vulnerabilities and agentic risk

Overview

This skill is purpose-built for UKC sandboxing, but it gives an agent cloud provisioning, remote execution, and file-sync authority with weak activation and SSH trust safeguards.

Install only if you intend to let the agent use your UKC account to create and delete sandbox VMs and sync selected local directories to a remote sandbox. Use a least-privilege UKC token, avoid syncing directories containing secrets, review files copied back before executing them locally, and prefer a version that pins SSH host keys and requires explicit confirmation for provisioning and sync steps.

SkillSpector

Vulnerability Patterns
  • Excessive AgencyUnrestricted Tool Access, Autonomous Decision Making, Scope Creep
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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
Findings (4)

Lp3

Medium
Category
MCP Least Privilege
Confidence
91% confidence
Finding
The skill clearly instructs the agent to invoke shell, bash, ssh, curl-like API interactions, and Node scripts, yet it does not declare corresponding permissions. That creates a capability/permission mismatch where reviewers or enforcement systems may underestimate what the skill can do, especially since it also handles inherited secrets and remote execution in a sandbox context.

Context-Inappropriate Capability

Medium
Confidence
88% confidence
Finding
The script provisions not only SSH access on port 2222 but also an additional HTTPS service mapping external 443 to internal 8080 without any clear need in the stated purpose of creating an isolated execution sandbox. Unnecessary exposed services expand the attack surface, may make the sandbox reachable over the web, and can permit unintended data exposure or remote interaction if workloads bind to port 8080.

Vague Triggers

Medium
Confidence
80% confidence
Finding
The trigger condition is broad: it activates on generic phrases like running code in a sandbox or any case where isolation is necessary. Over-broad activation can cause the skill to run in situations the user did not clearly intend, leading to unnecessary shell execution, sandbox creation, secret use via inherited environment variables, and remote sync of local files.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
This is a real vulnerability: disabling SSH host key verification allows a man-in-the-middle attacker to impersonate the sandbox endpoint and serve arbitrary files during the rsync pull. In this skill's context, the script is specifically used to retrieve data from an isolated sandbox, so trusting an unauthenticated remote host undermines the integrity of artifacts copied back to the local machine and can facilitate follow-on compromise or data poisoning.

Static analysis

No suspicious patterns detected.