keyboard-navigation-review

Pass

Audited by Gen Agent Trust Hub on Sep 5, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The workflow in SKILL.md instructs the agent to use the a11y-agent CLI tool for performing keyboard accessibility reviews.
  • [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from external websites and journey files to drive the testing process. This creates a surface where malicious instructions on a target page could potentially influence the agent's actions. * Ingestion points: Target URLs and journey files processed in SKILL.md. * Boundary markers: Explicit instructions in SKILL.md limit the agent to specific actions and forbid arbitrary JavaScript. * Capability inventory: Shell command execution via the a11y-agent tool. * Sanitization: Relies on natural language constraints defined in the workflow and reference files.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 5, 2026, 03:56 PM