ClawHub
Back to skill

Security audit

Prompt Benchmark

Security checks across malware telemetry and agentic risk

Overview

This skill is a coherent prompt-evaluation helper that uses bundled reference files and a local lint script without hidden persistence, credential access, or destructive behavior.

Install this if you are comfortable with the agent reading the skill's bundled rubric files and any prompt files you ask it to evaluate. The main operational consideration is that prompt content you provide will be analyzed locally by the skill workflow; I did not find evidence of hidden credential access, persistence, destructive actions, or automatic external model calls.

SkillSpector

Vulnerability Patterns
  • 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
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (1)

Lp3

Medium
Category
MCP Least Privilege
Confidence
90% confidence
Finding
The skill instructs the agent to read local reference files such as `references/scoring-rubric.md`, `references/report-schema.md`, and `references/model-compatibility.md`, which is a file-read capability. Because no permissions are declared, the runtime may not enforce least-privilege controls or may surprise operators about what the skill can access. In this skill’s context the reads appear limited to bundled reference material, so the risk is primarily governance and over-broad capability, not obvious data exfiltration.

VirusTotal

65/65 vendors flagged this skill as clean.

View on VirusTotal

Static analysis

No suspicious patterns detected.