aria-meter-name

Pass

Audited by Gen Agent Trust Hub on Jul 31, 2026

Risk Level: SAFE
Full Analysis
  • [SAFE]: The skill performs accessibility audits on HTML code. No evidence of credential exfiltration, malicious command execution, persistence mechanisms, or obfuscation was found.
  • [PROMPT_INJECTION]: The skill identifies a surface for indirect prompt injection because it requires the agent to ingest and analyze untrusted external HTML content.
  • Ingestion points: The agent is instructed to review rendered HTML and interactive components (defined in SKILL.md).
  • Boundary markers: There are no explicit instructions or delimiters provided to ensure the agent ignores natural language instructions that might be embedded within the audited HTML comments or attributes.
  • Capability inventory: The skill utilizes the agent's ability to perform code review and identify specific architectural patterns.
  • Sanitization: No sanitization or validation logic is defined to filter the input data before processing.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 31, 2026, 05:57 PM
Security Audit — agent-trust-hub — aria-meter-name