The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill retrieves data from WhatsApp messages sent by external contacts, which could contain instructions aimed at manipulating the agent's behavior.\n
Ingestion points: Functions in whatsapp_skill.js such as cmdMessages, cmdChats, and cmdSearch pull message bodies from external sources into the agent's context.\n
Boundary markers: The instructions do not define delimiters or specific 'ignore' rules for the retrieved data, increasing the likelihood that the agent might follow embedded instructions.\n
Capability inventory: The skill provides capabilities to send messages and manage chat sessions, which could be abused if an indirect injection is successful.\n
Sanitization: There is no automated filtering or sanitization of incoming message content to neutralize potential injection payloads.\n- [COMMAND_EXECUTION]: The skill instructs the agent to execute a local Node.js script for its operations.\n
All functional commands are implemented by calling node whatsapp_skill.js with various arguments.\n- [EXTERNAL_DOWNLOADS]: The skill requires downloading external dependencies during the initial setup.\n
Downloads and installs whatsapp-web.js and qrcode-terminal from the npm registry to enable WhatsApp Web automation.

whatsapp-skill