Back to skill

Security audit

Phantra AI: Ad Formats

Security checks for vulnerabilities and agentic risk

Overview

The skill provides a coherent ad-making workflow with expected photo uploads and explicit approval before paid rendering.

Reference photos are supplied to Phantra, potentially including faces or private rooms. Use images you intend to share and review the service's storage terms for sensitive material. Paid rendering requires your approval after a quote. The frontmatter formatting may need correction before the skill loads reliably.

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 instructs uploading user photos without a privacy disclosure.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The skill directs the agent to use add_element or upload_link to supply user pictures, potentially including faces and private rooms, to Phantra. It does not warn that these images will be transferred to the service; the later rendering-price confirmation only addresses spending credits.

Content

No source excerpt is available for this finding.