e2e-testing-expert
Warn
Audited by Gen Agent Trust Hub on Feb 16, 2026
Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [Indirect Prompt Injection] (MEDIUM): The skill's primary function involves navigating to and processing untrusted external web content via
page.goto(). This establishes an attack surface where malicious web pages could attempt to influence the agent's reporting or behavior. Evidence: Ingestion occurs inLoginPage.tsand test blocks; no boundary markers or sanitization are specified for external page content. - [Remote Code Execution] (MEDIUM): The script
scripts/generate-pom.tsperforms dynamic code generation (scaffolding) based on a URL's accessibility tree. This is a Category 10 risk where untrusted external structure could potentially be used to inject malicious logic into the generated POM classes. - [Command Execution] (LOW): The skill requires execution of shell commands such as
bun x playwright testand local utility scripts likescripts/analyze-traces.sh. This is standard for testing workflows but represents local execution. - [External Downloads] (LOW): The skill relies on standard, reputable Node.js packages including @playwright/test and @axe-core/playwright, typically fetched via the bun/npm registry.
Audit Metadata