Back to skill

Security audit

Phantra AI: Write Scene

Security checks for vulnerabilities and agentic risk

Overview

The skill writes and saves video scenes, with explicit user approval required before spending rendering credits.

Use photo references you intend to share with the service, review scene revisions, and approve only the rendering prices you accept. Independent review of the complete artifact found the scanner's scope concern mitigated by explicit approval boundaries; the supplied static scan reported no findings. This review does not verify the external service's implementation.

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 skill instructs pricing and rendering despite explicitly excluding both from its scope.

Medium
Category
Excessive Agency
Confidence
96% confidence
Finding

The manifest at L003 explicitly excludes pricing and rendering, assigning them to price-and-render. However, L087-L092 prescribe calling quote, approving a render with signed tokens, and polling the resulting job, extending this scene-writing skill into a credit-spending rendering workflow. Requiring user confirmation mitigates unauthorized spending but does not resolve the scope mismatch.

Content

No source excerpt is available for this finding.