Back to skill

Security audit

Generative Engine Optimization (GEO)

Security checks for vulnerabilities and agentic risk

Overview

The skill is marked suspicious because scanners report that it sends customer-derived prompts to external AI assistants without a privacy warning; this is a privacy concern, not evidence of malicious intent.

Before installing, consider whether you are comfortable sending customer-derived prompts to external AI providers. Review and sanitize prompts to remove personal, identifying, or confidential information, and use approved providers and accounts. The scan did not report malicious behavior, but coverage is incomplete because one local path-like reference could not be resolved.

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
Vulnerability Patterns
  • Output HandlingUnvalidated Output Injection, Cross-Context Output, Unbounded Output
  • 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)

The workflow sends customer-derived prompts to multiple external AI assistants without a privacy warning.

Medium
Category
Output Handling
Confidence
94% confidence
Finding

The workflow requires prompts derived from customer interviews, support tickets, or search terms to be run across multiple external AI assistants, without first requiring privacy review or sanitization. If customer wording contains personal information or confidential business details, an executing agent could disclose that information to third-party providers with differing retention policies. The legitimate benchmarking purpose and buyer-question framing reduce suspicion of malicious intent, but the skill's public-fact requirements for published content do not protect these input prompts.

Content

No source excerpt is available for this finding.