local-business-website-demo-cold-outreach
Pass
Audited by Gen Agent Trust Hub on Oct 31, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to visit, audit, and scrape data from external local business websites. This content is subsequently used to generate website outlines and prompts, creating a surface for potential injection attacks.
- Ingestion points: The agent retrieves performance data, design issues, and contact information (email, social links, WhatsApp) from arbitrary external URLs into its processing context (Step 2 and Step 6).
- Boundary markers: The instructions lack specific delimiters or safety prompts to tell the agent to ignore hidden instructions that might be present in the HTML or metadata of the target websites.
- Capability inventory: The agent uses retrieved data to generate structured outlines and prompts intended for use in other LLMs (Grok, ChatGPT, Claude).
- Sanitization: There is no logic provided to sanitize, filter, or validate the content retrieved from external websites before it is processed.
Audit Metadata