ClawHub
Back to skill

Security audit

6skill

Security checks across malware telemetry and agentic risk

Overview

This PRD skill is mostly documentation, but it includes an unsafe unattended coding-agent workflow that can repeatedly change a project without permission prompts.

Install only if you intend to use it as PRD documentation and review the agent workflow carefully. Do not run the unattended permission-skipping loop; run one story at a time in a branch or worktree, keep permission prompts enabled, inspect diffs before commits, and verify the publisher/package identity because the metadata is inconsistent.

SkillSpector

By NVIDIA
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 (2)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The document asserts that the skill only edits PRDs, but the actual guidance instructs agents to implement code, create branches, run checks, commit changes, and update auxiliary files. This mismatch can mislead users and safety systems about the skill’s effective capabilities, increasing the chance that repository-modifying automation is run with insufficient review or controls.

Missing User Warnings

Medium
Confidence
97% confidence
Finding
The unattended loop invokes the agent with `--dangerously-skip-permissions`, enabling repeated autonomous code changes without permission prompts or an adjacent warning about the risks. In this context, that creates a realistic path to unreviewed repository modification, destructive commands, secret exposure, or persistence of bad changes across iterations.

VirusTotal

64/64 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.