The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface for indirect prompt injection by ingesting untrusted data from an external messaging service (WhatsApp).
Ingestion point: references/webhook-route.ts receives POST requests containing message content via the wapi service.
Boundary markers: The reference implementation lacks explicit delimiters or instructions to ignore embedded commands within the message body.
Capability inventory: The skill provides extensive tools for interaction including sending messages (sendMessage), uploading files (upload), and reacting to messages (react) in references/wapi-client.ts. If an agent is configured to automate actions based on incoming WhatsApp messages, these capabilities could be misused via injection.
Sanitization: No specific content sanitization or validation is demonstrated in the route handler before data is logged or processed.
[SAFE]: The skill utilizes vendor-specific infrastructure belonging to the author (crafter-station) for API endpoints and documentation, such as api.wapi.crafter.run and wapi.crafter.run.
[SAFE]: The skill provides security-hardened reference code, specifically using import "server-only" to prevent accidental credential leakage to client-side bundles and implementing timingSafeEqual for cryptographic signature verification in webhooks.

wapi-nextjs