accessibility
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTION
Full Analysis
- Command Execution: The skill uses Python's subprocess modules to drive Node.js, the npx package runner, and PowerShell scripts. These operations are used to execute the scanner engine and manage system-level window focus on Windows to ensure accurate screen reader capture.
- External Downloads: The scanner CLI fetches the version-pinned
@axe-core/clipackage from the public npm registry during its first run. This is part of the tool's core functionality for performing accessibility audits. - Indirect Prompt Injection Surface: The skill ingests data from human-authored design intent records and external accessibility findings. This processing surface is mitigated by extensive schema validation, semantic checks, and a recursive privacy guard that rejects credential-shaped content.
- Dynamic Execution Patterns: The harness dynamically imports internal probe modules and injects observation scripts into Playwright browser contexts. These mechanisms are used to monitor live page behaviors, such as ARIA live region announcements, and are scoped to the skill's distributed scripts.
Audit Metadata