The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data from the Creem API and webhooks, including customer-controlled fields such as metadata and custom_fields. This creates an attack surface where malicious instructions could be embedded in transaction data for an agent to process later.
Ingestion points: Data enters the agent's context through API responses from endpoints like /v1/checkouts and /v1/customers, as well as through webhook event payloads (e.g., checkout.completed) documented in WEBHOOKS.md.
Boundary markers: The skill does not currently instruct the agent to use specific delimiters or provide warnings to ignore embedded instructions when interpolating these external fields into prompts.
Capability inventory: The skill allows the agent to perform network requests (Checkouts/Subscriptions API) and potentially influences database operations based on this external data.
Sanitization: While the code examples show standard integration logic, there is no explicit guidance for the agent to validate or sanitize user-supplied metadata strings before further action.
[SAFE]: The skill demonstrates a high level of security awareness by explicitly instructing developers to use crypto.timingSafeEqual for webhook signature verification to prevent timing attacks and ensuring that API keys are never exposed in client-side code.

creem-api