Back to skill

Security audit

four-layer-photo-poster

Security checks for vulnerabilities and agentic risk

Overview

This skill is a disclosed photo-poster workflow with a local assembly script and no evidence of hidden access, persistence, or data exfiltration.

Install only if you are comfortable providing photos to the image-generation workflow and producing derived stylized layers. Review generated outputs for unwanted invented details, but the artifact does not show hidden security-impacting behavior.

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)

Tp4

High
Category
MCP Tool Poisoning
Confidence
95% confidence
Finding

The code does perform exact four-band assembly on a 3:4 vertical canvas, which partially matches the description. However, the declared purpose says each supplied photo is turned into its own poster and implies a continuous source scene across the four stages. In reality, the script accepts four specifically named image inputs and stitches them into a single poster, with each band coming from a different file. There is no logic to derive multiple stages from one source photo or preserve/construct a continuous scene across bands. Thus the primary behavior is materially different from the declared description.

Content

No source excerpt is available for this finding.