marketing-email
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill provides instructions to install the 'agent-browser' utility from Vercel Labs' official GitHub repository (github.com/vercel-labs/agent-browser). This is a well-known service and the reference is for legitimate setup purposes.
- [COMMAND_EXECUTION]: The shared patterns reference file contains shell commands for environment setup, including 'npm install' for the agent-browser CLI and 'npx playwright install chromium'. These commands are standard for installing developer tools.
- [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface in 'references/shared-patterns.md' under 'Context C — Live Website URL Audit'. It is designed to ingest data from user-provided live URLs to perform audits.
- Ingestion points: Content from live URLs provided by the user for audits (references/shared-patterns.md).
- Boundary markers: The instructions do not define specific delimiters or instructions to ignore embedded directives in the audited web content.
- Capability inventory: The skill has file-write capabilities, saving deliverables to the './brands/' directory (SKILL.md sections 13, 14, 15).
- Sanitization: No explicit sanitization or validation of the ingested web content is performed before processing.
Audit Metadata