Back to skill

Security audit

vibe-app-quality-checker

Security checks for vulnerabilities and agentic risk

Overview

The app-review skill needs review because it directs testing of potentially consequential jobs, including bulk messaging, without requiring an isolated test environment.

Use an isolated staging app with synthetic data, test recipients, mocked external services, and bounded workloads. Explicitly authorize any real messaging or other consequential operations; otherwise use code inspection and report runtime behavior as untested. The full scanner report was complete, and all supplied artifact files were reviewed.

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 the agent to trigger heavy tasks without warning about potentially consequential side effects.

Medium
Category
Output Handling
Confidence
88% confidence
Finding

The workflow instructs the agent to trigger heavy tasks, explicitly including sending many messages, without requiring confirmation that execution is isolated or authorized. Such testing could send actual notifications, incur service charges, or consume substantial resources even with synthetic data. The prohibition on real user data and real payments reduces risk but does not cover these other consequential effects or guarantee that a demo environment is isolated.

Content

No source excerpt is available for this finding.