The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill instructs the agent to read the Vercel CLI authentication token from the local filesystem path ~/Library/Application Support/com.vercel.cli/auth.json to facilitate API-based deployment.
[COMMAND_EXECUTION]: Automated setup involves multiple shell commands and API requests using curl and npx to Vercel, Kapso, and Maton for provisioning infrastructure and webhooks.
[PROMPT_INJECTION]: The skill creates an automated bot susceptible to indirect prompt injection from external WhatsApp messages or the public test endpoint.
Ingestion points: User-provided message content via the webhook in templates/app/api/whatsapp/webhook/route.ts and the test endpoint in templates/app/api/test/route.ts.
Boundary markers: Absent. The system instructions in templates/lib/prompts.ts lack delimiters or specific instructions to separate untrusted user input from core system instructions.
Capability inventory: Integrated tools in templates/lib/tools.ts allow for the creation, modification, and deletion of Google Calendar events and Zoom meetings.
Sanitization: External input is passed directly to the model without filtering, escaping, or validation.

whatsapp-agent