The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill contains a hardcoded API credential string (pk_CzEKBTWFiuNLgz4wciLsS59n) assigned to the publishableKey property in the Quickstart code snippet.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an attack surface where external or untrusted data could be interpolated into API request parameters or bodies, potentially leading to indirect prompt injection if not properly sanitized.
Ingestion points: Arguments for SDK resource methods (e.g., client.products.list) and the low-level client.request method described in SKILL.md.
Boundary markers: There are no explicit boundary markers or instructional guards defined in the documentation to separate user data from API command structures.
Capability inventory: The skill enables the agent to perform network requests to external endpoints (the configured baseUrl) using the SDK's fetch-based implementation.
Sanitization: The skill provides and encourages the use of Zod schemas (@spree/sdk/zod) for validating external data and signature verification for webhooks (verifyWebhookSignature), which serve as relevant security mitigations.
[EXTERNAL_DOWNLOADS]: The skill instructs the installation of external NPM packages.
Evidence: npm install @spree/sdk and npm install @spree/admin-sdk@next.
Context: These packages are official vendor resources associated with the skill author ('spree') and represent standard dependency management for this SDK.

spree-typescript-sdk