The Agent Skills Directory

[DATA_EXFILTRATION]: The skill communicates exclusively with the official Meta Graph API (graph.facebook.com) to perform messaging operations as intended. Placeholders for credentials are provided in .env.example files, and the utility scripts include logic to mask secrets in console output to prevent accidental exposure.
[COMMAND_EXECUTION]: The provided scripts (setup_project.py, validate_config.py) are development utilities designed to initialize projects and test connectivity. They do not execute arbitrary shell commands based on untrusted input.
[PROMPT_INJECTION]: An assessment of the instructional content and documentation found no attempts to override agent behavior, bypass safety filters, or extract system prompts.
[INDIRECT_PROMPT_INJECTION]: The skill establishes an ingestion point for untrusted data via webhooks which receive user messages from WhatsApp. While this surface area exists, the skill provides the necessary security mechanisms (HMAC-SHA256 validation in webhook-handler.ts and webhook_handler.py) to authenticate the source. The documentation further guides developers on implementing system prompts for AI integrations.
Ingestion points: Webhook POST handlers in assets/boilerplate/python/app.py and assets/boilerplate/nodejs/src/index.ts.
Boundary markers: Documentation in references/automation-patterns.md recommends system prompts for context separation.
Capability inventory: The skill can send various message types (text, media, interactive) via the WhatsApp API.
Sanitization: Standard message extraction is implemented in the boilerplate, with developers instructed to handle content logic.

whatsapp-cloud-api