The Agent Skills Directory

[COMMAND_EXECUTION]: The script scripts/message_router.py utilizes __import__("time") for dynamic module loading. This is used strictly to retrieve current timestamps for session activity tracking and does not facilitate arbitrary code execution.
[CREDENTIALS_UNSAFE]: API tokens and Meta application secrets are managed through environment variables and Pydantic settings, as demonstrated in app/config.py and scripts/webhook_handler.py. No hardcoded credentials or sensitive keys were identified in the source files.
[DATA_EXFILTRATION]: Network operations are directed solely to the official Meta Graph API (graph.facebook.com) to support the skill's messaging functionality. No unauthorized data transmission to third-party or suspicious domains was observed.
[EXTERNAL_DOWNLOADS]: The skill documentation lists standard, well-maintained Python libraries such as fastapi, httpx, and redis. These dependencies are common in production environments and do not pose a supply chain risk in this context.
[PROMPT_INJECTION]: The skill acts as a gateway for untrusted user input from WhatsApp, creating a potential surface for indirect prompt injection. This is a standard risk for chatbot infrastructure and is documented here for completeness.
Ingestion points: Incoming message content is parsed in scripts/webhook_handler.py and routed in scripts/message_router.py.
Boundary markers: No specific delimiters are used to isolate user-provided text before it is forwarded to backend agents.
Capability inventory: The skill can trigger outbound messages via Meta's API and update conversation state in a Redis database.
Sanitization: No content-based filtering or sanitization of incoming text is performed.

whatsapp-cloud-api-agent