The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements logic to process user roles and permissions extracted from request metadata (req.user.klaviyoRole). This creates a theoretical surface for indirect prompt injection where a maliciously crafted role or identity attribute could attempt to influence the agent's logic or bypass application-level checks.
Ingestion points: req.user.klaviyoRole is used in references/implementation.md and references/examples.md to determine permissions.
Boundary markers: The implementation uses a strict AppRole enum and a static ROLE_PERMISSIONS map to delimit allowed actions, which acts as a robust boundary.
Capability inventory: The skill facilitates interactions with the Klaviyo API via fetch calls to official endpoints and uses scoped ApiKeySession instances.
Sanitization: The code validates provided roles against the static configuration and is designed to "fail closed" if an unrecognized role is encountered.

klaviyo-enterprise-rbac