marketing-referral
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references and provides instructions for installing the
agent-browserdependency from the Vercel Labs GitHub repository. - Evidence:
npx skills add https://github.com/vercel-labs/agent-browser --skill agent-browserinreferences/shared-patterns.md. - Note: Vercel Labs is a recognized service provider, and the reference is documented neutrally as a setup step.
- [COMMAND_EXECUTION]: The skill contains shell commands intended for the user to set up their local environment and install necessary tools.
- Evidence:
npm install -g agent-browser && npx playwright install chromiuminreferences/shared-patterns.md. - [PROMPT_INJECTION]: The skill's workflow includes auditing live website URLs, which presents a surface for indirect prompt injection (Category 8).
- Ingestion points: The
Live Website URL Auditmode described inreferences/shared-patterns.mdand referenced inSKILL.md. - Boundary markers: The skill does not explicitly define markers to separate external web content from the agent's instructions.
- Capability inventory: The skill performs file system writes (saving deliverables to
./brands/{brand-slug}/campaigns/referral/) and web browsing. - Sanitization: No explicit content sanitization or instruction-filtering logic is provided for the audited data.
Audit Metadata