frontend-rails
Pass
Audited by Gen Agent Trust Hub on Nov 12, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill downloads the
playwrightpackage and its dependencies during setup to provide browser automation capabilities. - [COMMAND_EXECUTION]: Utility scripts such as
compare.mjsandwalk.mjsexecute Node.js commands to run internal auditing logic viaspawnandspawnSync. - [INDIRECT_PROMPT_INJECTION]: The skill exposes a surface for indirect prompt injection through its auditing and walkthrough features:
- Ingestion points:
gate/walk.mjsscrapes text and control labels from web pages, andgate/run.mjscaptures screenshots for AI analysis. - Boundary markers: Subagent prompts in
gate/walkthrough.mdandgate/vision-questions.mddo not include explicit instructions to disregard embedded malicious text within the audited page content. - Capability inventory: The subagent is empowered to interact with the browser (clicking, typing, selecting) based on the instructions it interprets from the page.
- Sanitization: No sanitization is performed on the text extracted from the pages before being sent to the AI subagent.
Audit Metadata