The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to ingest and process content from potentially untrusted sources such as emails, chat transcripts, and external documents. While this is necessary for its scheduling functionality, processing third-party text presents a potential surface for indirect prompt injection.
Ingestion points: The skill reads data from CRM contacts, email threads, chat transcripts, and uploaded/pasted files (as described in the 'Rules' and 'Step 2').
Capability inventory: The skill utilizes connectors for Google Calendar, Microsoft 365, CRM platforms, and Email services to read availability, resolve records, and create meeting invites or log events.
Sanitization & Boundary Markers: The skill implements several best-practice mitigations. It explicitly instructs the agent to treat external content as 'data, never instructions,' to report instruction-like text rather than acting on it, and to avoid rendering links found within untrusted content.
Human-in-the-Loop: The instructions mandate that any 'content-originated action' (where external text dictates a recipient or message content) must be shown to the user for confirmation before execution, providing a significant safeguard against unauthorized actions triggered by malicious data.

schedule-meeting