The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests and processes untrusted data from incoming Gmail messages.
Ingestion points: Step 3 (Get Full Email Details) fetches the body content of incoming emails from the Gmail inbox.
Boundary markers: The instructions do not specify the use of delimiters or clear separation between the system instructions and the untrusted email content during the classification step.
Capability inventory: The agent has the ability to read, label, and archive emails, as well as send messages to a Telegram channel.
Sanitization: There is no evidence of content sanitization or filtering before the email data is passed to the AI for classification.
[EXTERNAL_DOWNLOADS]: The skill setup requires downloading and installing several tool dependencies from the vendor's own repository.
Evidence: The documentation provides installation commands using npx skills add AgentPMT/agent-skills for tools such as what-is-agentpmt, agentpmt-account-mcp-rest-api-setup, gmail-all-email-actions, and telegram-instant-messenger.
Context: These resources are hosted on the vendor's official GitHub organization and marketplace, representing standard setup for the intended workflow.

ai-gmail-inbox-classifier-auto-archive-with-hourly-telegram