redesign-inventory
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill operates by analyzing live application interfaces, including DOM trees and accessibility snapshots. This data ingestion path is susceptible to indirect prompt injection if the application being audited contains malicious instructions within its UI elements (e.g., button labels, aria-labels, or hidden text). * Ingestion points: accessibility snapshots and DOM query outputs. * Capability inventory: The skill can navigate the browser, click elements, and monitor network traffic. * Boundary markers: No explicit markers are defined to isolate data from instructions. * Sanitization: There is no evidence of text sanitization or filtering before the data is processed by the agent.
- [EXTERNAL_DOWNLOADS]: The skill instructions recommend installing an external dependency via 'npx skills add stablyai/orca'. This involves downloading code from an external repository to provide necessary browser-driving capabilities.
- [COMMAND_EXECUTION]: The documentation includes shell commands for environment configuration and tool management using the 'npx' package runner.
- [DYNAMIC_EXECUTION]: The skill provides JavaScript templates for the agent to execute within the browser context to audit DOM properties and control states. These snippets are used to measure layout and verify functionality.
Audit Metadata