complete-link-purchase
Pass
Audited by Gen Agent Trust Hub on Nov 6, 2026
Risk Level: SAFE
Full Analysis
- Instructional Security Guardrails: The skill incorporates strong instructions for the agent to prevent the exposure of sensitive payment information. It explicitly forbids the agent from repeating card details in chat or persisting them in files or logs, which is a key security practice.
- Mitigation of Indirect Prompt Injection: The skill includes a dedicated section on treating merchant content as data rather than instructions. This provides a layer of protection against potential malicious directives that might be embedded in third-party websites during the checkout process.
- Defined Operational Limits: The skill outlines specific financial and temporal limits, such as a $5,000 transaction cap and specific expiration windows, which provide necessary boundaries for the agent's autonomous actions.
Audit Metadata