ClawHub
Back to skill

Security audit

Security code review

Security checks across malware telemetry and agentic risk

Overview

This is a security-review guidance skill that inspects code and reports vulnerabilities, with only a minor disclosure issue around optional local report files.

Install this for repositories you want an agent to inspect for security issues. Review any generated chat report or `.shield_security/` files as sensitive because they may include real vulnerabilities, file paths, or discovered secrets, and prefer explicit approval before saving reports to the workspace.

SkillSpector

By NVIDIA
Vulnerability Patterns
  • 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
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (2)

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The skill claims security analysis must use only read-only tools, but immediately allows creation of artifacts in the workspace. That contradiction can normalize unauthorized file writes during a supposedly non-mutating review flow, increasing the chance an agent modifies the repository or leaves persistent data without clear user consent.

Missing User Warnings

Medium
Confidence
91% confidence
Finding
The markdown permits writing artifacts to `.shield_security/` during security analysis but does not provide a clear upfront warning that the workspace may be modified. In an agentic setting, this can mislead users into expecting a read-only review while the skill persists files locally, which may overwrite data, leak sensitive findings into the repo, or violate least-privilege expectations.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

Detected: suspicious.exposed_secret_literal

File appears to expose a hardcoded API secret or token.

Critical
Code
suspicious.exposed_secret_literal
Location
SKILL.md:37