Security audit
Security checks for vulnerabilities and agentic risk
This food-ordering skill is mostly coherent, but it asks to retain meal/order history and includes broad local command execution that is not well scoped.
Install only if you are comfortable with the agent using delivery websites and saving food preferences and order history. Avoid granting local command execution unless the publisher narrows it to specific, necessary scripts and adds clear user consent for stored data.
The skill uses broad, everyday phrasing around a common consumer task, which can make invocation boundaries ambiguous and increase the chance the skill activates in situations the user did not clearly intend. In a browser-enabled ordering skill, accidental or over-broad activation can lead to unnecessary data collection, navigation to third-party sites, or unintended assistance with purchases.
The skill explicitly stores user preferences and order history in memory files but does not present a clear user-facing notice, consent flow, retention policy, or deletion mechanism. Because these records can reveal location, dietary restrictions, habits, and spending patterns, silent retention creates privacy risk and potential exposure of sensitive personal data.
The skill declares exec for local script execution even though its stated purpose is food discovery and ordering, and the document provides no narrowly scoped justification, constraints, or allowed commands. Unnecessary local execution materially expands the attack surface: if later prompt injection or tool misuse occurs, the agent could run arbitrary local scripts, access local files, or chain into broader system compromise.
