ClawHub
Back to skill

Security audit

Code Delegate Tool Free

Security checks across malware telemetry and agentic risk

Overview

This skill is a coding delegation helper, but it recommends broad local command execution with automatic file-edit approval, which needs careful review before installation.

Install only if you intentionally want an agent to delegate coding tasks to a locally authenticated Claude CLI. Use it in a disposable or isolated project directory, avoid permission-bypass modes, review generated changes before trusting them, and do not point it at repositories or directories containing secrets or sensitive data.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
Findings (3)

Intent-Code Divergence

High
Confidence
98% confidence
Finding
The skill explicitly recommends running the delegated code CLI with `--permission-mode bypassPermissions`, which suppresses normal confirmation controls for file edits and command execution. In a tool that delegates arbitrary programming tasks to a local CLI, this materially increases the risk of unauthorized file modification, destructive shell actions, or execution of unsafe model-suggested commands without user review.

Vague Triggers

Medium
Confidence
87% confidence
Finding
The trigger condition is broad enough to match many ordinary development conversations, which can cause the skill to activate unexpectedly and invoke `exec` for tasks the user did not explicitly intend to delegate. In a skill that can launch a local code CLI and modify project files, accidental activation raises the likelihood of unreviewed command execution and unintended workspace changes.

Missing User Warnings

High
Confidence
97% confidence
Finding
The documentation normalizes automatic permission acceptance for delegated code execution but does not prominently warn users that the CLI may edit files or run commands with limited oversight. In the context of a local execution skill, this creates a strong risk of users authorizing impactful actions without understanding that the delegated tool can alter source code, run tests, install dependencies, or execute harmful shell operations.

VirusTotal

VirusTotal findings are pending for this skill version.

View on VirusTotal

Static analysis

No suspicious patterns detected.