Back to skill

Security audit

Functional-Food Officers

Security checks for vulnerabilities and agentic risk

Overview

The skill provides user-controlled business workflows, though its food-marketing examples need independent human review.

Provide only the business records needed for your task and retain approval before publishing or taking business actions. Have a qualified reviewer check health-related wording and applicable regulations; the sample compliance checks are not reliable proof that marketing copy is suitable for publication.

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
  • 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 (1)

The outward-facing content example includes physiological mechanisms despite an explicit internal-only restriction.

Medium
Category
Excessive Agency
Confidence
97% confidence
Finding

The public-facing draft describes gel formation, delayed carbohydrate absorption, and slower post-meal blood glucose increases, contradicting the task’s explicit restriction that physiological mechanisms remain internal. In a food-marketing skill, this example could steer generated promotional copy toward unsupported product-specific health implications, particularly because its self-check incorrectly reports no claims and cites no supporting test data. This is a content-safety and compliance-control failure, not evidence of malicious instruction injection.

Content

No source excerpt is available for this finding.