ClawHub
Back to skill

Security audit

AI Agent Orchestration Advisor

Security checks for vulnerabilities and agentic risk

Overview

This skill is an advisory guide for choosing and designing multi-agent systems, with disclosed example code users may adapt but no packaged code that runs by itself.

Before using the generated examples in a real environment, review any web-search or third-party tools, replace them with approved internal services where required, keep secrets in environment variables, and follow the skill's own data-minimization and human-review guidance.

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
  • 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
Category
Not specified by scanner
Confidence
95% confidence
Finding

The skill explicitly states it will not make external network calls, yet the provided runnable scaffold includes SerperDevTool() and a crew.kickoff(...) flow that is designed to trigger web search when executed by a user. This mismatch can mislead users into treating the generated code as compliant with a no-network boundary, increasing the risk of unintended data egress or policy violations in restricted environments.

Content

No source excerpt is available for this finding.

Intent-Code Divergence

Low
Category
Not specified by scanner
Confidence
81% confidence
Finding

The manifest scope at L004-L005 emphasizes that the skill does not run, install, or deploy frameworks, yet the example interactions at L357 and L361 describe generating a complete scaffold with concrete tools and providing a LangGraph code fix with a PostgreSQL checkpointer. While this may still be advisory, the wording blurs the line between high-level architecture advice and implementation-level operational guidance.

Content

No source excerpt is available for this finding.

Static analysis

No suspicious patterns detected.