The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses a pre-installed shopify binary to interact with Shopify's official Model Context Protocol (MCP) server. It executes commands such as shopify status, shopify authorize-url, and shopify call-tool to manage store resources.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external sources, including Shopify store data (products, orders, customers) and official documentation from shopify.dev.
Ingestion points: External data enters the context via tool calls like search_products, list-orders, and search_docs_chunks (documented in SKILL.md).
Boundary markers: The instructions do not define explicit text delimiters for external content, but they emphasize inspecting tool-level results separately from command status.
Capability inventory: The skill has the capability to execute the shopify binary and perform write operations like create-product, create-discount, and import-mock-shop-catalog (documented in manifest.yaml).
Sanitization: The manifest.yaml defines a default: ask policy for all write methods, ensuring a human-in-the-loop requirement for operations that modify data.
[PROMPT_INJECTION]: The instructions contain directives to filter output, such as sharing only the connect_url from the authorization response. While identified as a concealment pattern by static analysis, this is a standard security practice in OAuth flows to prevent leaking sensitive technical parameters or internal tokens to the user.

shopify