The Agent Skills Directory

[SAFE]: The skill demonstrates secure handling of provider credentials by using environment variables (e.g., TWILIO_ACCOUNT_SID, WHATSAPP_ACCESS_TOKEN) rather than hardcoding secrets in the implementation logic.
[SAFE]: Secure OTP management is implemented by hashing passcodes with bcrypt and enforcing short-lived expiration and attempt limits, which protects against local data exposure.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface where inbound messaging content is processed and potentially interpolated into templates. This is an inherent risk factor for messaging skills but is documented here as an observation.
Ingestion points: ConsentManager.processInboundReply and WhatsAppWebhookHandler._process_message ingest untrusted user-provided message bodies from SMS and WhatsApp replies.
Boundary markers: No specific delimiters or instructions to the agent to ignore embedded commands in processed data were identified in the template render logic.
Capability inventory: The skill enables network operations via fetch, SMS delivery via Twilio/SNS SDKs, and persistent storage via database records.
Sanitization: The renderTemplate function performs simple string replacement for template variables without explicit escaping or validation of user-provided content.
[COMMAND_EXECUTION]: No instances of arbitrary command execution or unsafe shell patterns were found in the provided code snippets.
[EXTERNAL_DOWNLOADS]: The skill references standard, well-known libraries such as twilio, boto3, and phonenumbers from official registries. These are treated as safe dependencies.

backend-sms-messaging