ClawHub
Back to skill

Security audit

Agentic Advertising: Build Autonomous Media-Buying Agents

Security checks for vulnerabilities and agentic risk

Overview

This is a non-running guide, but its copyable examples can authorize real ad-budget wallet and escrow actions while its sandbox and credential guidance is inconsistent.

Install only if you intend to study or build GreenHelix-based ad-buying agents. Treat every wallet, escrow, release, cancellation, and budget-allocation example as potentially money-moving until you confirm the endpoint and account are sandbox-only. Use test accounts, hard spending caps, secret-manager or environment-based credentials, and explicit human approval gates before any production campaign spend.

SkillSpector

Vulnerability Patterns
  • Data ExfiltrationExternal Transmission, Env Variable Harvesting, File System Enumeration
  • MCP Tool PoisoningHidden Instructions, Unicode Deception, Parameter Description Injection
  • Prompt InjectionInstruction Override, Hidden Instructions, Exfiltration Commands
  • Privilege EscalationExcessive Permissions, Sudo/Root Execution, Credential Access
  • Supply ChainUnpinned Dependencies, External Script Fetching, Obfuscated Code
Findings (3)

Intent-Code Divergence

Medium
Confidence
91% confidence
Finding
The guide states that no API key is required to get started, but the setup code immediately introduces and uses a bearer API key. This mismatch can mislead operators about the authentication model, causing them to provision or paste sensitive credentials unnecessarily and undermining trust in the guide's security assumptions.

Intent-Code Divergence

Medium
Confidence
94% confidence
Finding
The skill metadata declares AGENT_SIGNING_KEY as the required credential, but the example code instead uses an undeclared API key. This discrepancy encourages insecure credential handling, because users may expose or hardcode additional secrets not covered by the skill manifest or platform permission model.

Missing User Warnings

Medium
Confidence
82% confidence
Finding
The skill repeatedly demonstrates escrow creation, release, cancellation, wallet operations, and budget allocation for autonomous ad spend, yet only provides a light educational notice at the top. In this context, insufficient warning is dangerous because users may treat the examples as low-risk demos when they actually model real financial commitments and external spending actions.

Static analysis

No suspicious patterns detected.