The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes data from external sources via web_search and search_policies. This represents a potential surface for indirect prompt injection, where external content might attempt to influence the agent's behavior.
Ingestion points: Untrusted data enters the agent context through web_search and store policy calls defined in SKILL.md.
Boundary markers: No specific delimiters or instructions to ignore embedded commands are present in the provided text.
Capability inventory: The skill is capable of reading data and recommending products; state-changing actions like writing to the cart or memory require explicit user confirmation.
Sanitization: No explicit sanitization or filtering of external content is described in the instructions.
User-Centric Controls: The skill explicitly restricts writing to the cart or memory unless the customer specifically asks, which aligns with security best practices for agent autonomy.

purchase-research