Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill utilizes npx @useclawlink/cli to fetch and execute the ClawLink command-line interface from the public NPM registry. This is the official tool provided by the vendor (ClawLink) to facilitate the integration.
- [COMMAND_EXECUTION]: The skill uses shell commands to invoke the npx utility for authenticating with the service and executing WhatsApp-related actions.
- [INDIRECT_PROMPT_INJECTION]: As the skill reads external data (such as message templates and profile information) from WhatsApp, there is an inherent surface for indirect prompt injection. The skill includes design safeguards by instructing the agent to always read data before acting and to obtain user confirmation before performing any write operations (e.g., sending messages or deleting templates).
- Ingestion points: Data is ingested through actions like whatsapp_get_message_templates and whatsapp_get_business_profile in SKILL.md.
- Boundary markers: The skill explicitly instructs the agent to 'Reads first; confirm with the user before any write'.
- Capability inventory: The skill has the capability to send messages (whatsapp_send_location, whatsapp_send_interactive_buttons) and delete resources (whatsapp_delete_message_template).
- Sanitization: No specific sanitization or filtering of external content is mentioned in the skill definition.
Audit Metadata