accessibility-testing
Pass
Audited by Gen Agent Trust Hub on Sep 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill exhibits an attack surface for indirect prompt injection as it is designed to ingest and process untrusted external data for accessibility auditing.
- Ingestion points: Untrusted content enters the agent's context through external URL auditing and HTML code snippets defined in the evaluation suite (evals/accessibility-testing.yaml).
- Boundary markers: The instructions do not specify explicit delimiters or 'ignore embedded instructions' prompts for the agent when processing the content of audited pages.
- Capability inventory: The skill utilizes powerful capabilities including browser automation (page.evaluate), shell command execution via local fleet tools (aqe, vibium), and agent coordination (FleetManager).
- Sanitization: No evidence of sanitization or filtering of the audited content is present before it is analyzed by the reasoning engine.
Audit Metadata