ClawHub
Back to skill

Security audit

opus

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed model-routing guide with no code, persistence, credential access, or hidden data movement, though its routing rules are somewhat subjective.

This skill is reasonable to install if you want cost-aware Claude model routing. Treat its rules as guidance rather than a strict safety policy, and manually choose a stronger model for ambiguous, high-impact, production, legal, financial, medical, or security-sensitive work.

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 routing rules are materially ambiguous: phrases like 'most user conversations' and 'if a human needs >30 seconds of focused thinking' are subjective, hard to operationalize, and likely to produce inconsistent model selection. In a security-relevant routing skill, this can cause systematic under-routing of complex or sensitive tasks to cheaper models, degrading analysis quality and increasing the chance of missed risks or incorrect outputs.

Vague Triggers

Medium
Confidence
92% confidence
Finding
The decision tree relies on broad categories like 'code, analysis, planning, writing, or multi-step,' which cover a large fraction of real-world prompts and do not meaningfully constrain when escalation should occur. This underspecification makes the skill unpredictable and easy to misapply, potentially routing sensitive or complex tasks incorrectly and undermining reliability, especially for security, debugging, or production decision-making.

Static analysis

No suspicious patterns detected.