ClawHub
Back to skill

Security audit

Model Verifier

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed model-checking checklist, but its identity tests are unreliable and should not be treated as proof of which model is running.

Install only if you want a lightweight model-authenticity checklist. Treat its results as rough observations, not proof: safety style and language-pattern claims can be inaccurate, and recorded Q&A may include conversation context you choose to provide.

A.I.G

Vulnerability Patterns
  • Skill Instruction HijackingAlters the agent's session goals or safety constraints when the skill loads
  • Agent Memory PoisoningWrites attacker-controlled rules into memory that affect later sessions
  • Remote Payload Retrieval and ExecutionFetches external code whose behavior can change after review
  • Embedded Malicious CodeShips malicious scripts inside the skill and executes them locally
  • Unauthorized Access and Privilege EscalationObtains permissions beyond the task's legitimate needs

SkillSpector

Vulnerability Patterns
  • Trigger AbuseOverly Broad Trigger, Shadow Command Trigger, Keyword Baiting Trigger
  • 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)

Vague Triggers

Medium
Confidence
95% confidence
Finding
The activation guidance is broad enough that the skill may run whenever a user asks to 'verify model' or check authenticity, without clear limits on consent, scope, or destination of collected outputs. In this skill, automatic invocation matters because the workflow instructs the agent to probe model identity through behavioral tests and record Q&A evidence, which can lead to unnecessary fingerprinting or policy-evasion-oriented evaluation.

Natural-Language Policy Violations

Medium
Confidence
97% confidence
Finding
The skill uses language-pattern heuristics to identify the underlying model, including claims that specific vendors 'think' in Chinese or English, without user opt-in or reliable justification. This is dangerous because it encourages covert fingerprinting based on inferred internal reasoning style, which is both privacy-sensitive and easily repurposed to tailor jailbreaks, evade safeguards, or make unsupported claims about hidden chain-of-thought behavior.

Static analysis

No suspicious patterns detected.