The Agent Skills Directory

[SAFE]: The skill uses official Google Calendar API endpoints (googleapis.com) for all operations. Authentication is handled through a secure connection selector mechanism (Zapier-managed OAuth or environment variables), avoiding hardcoded credentials.
[COMMAND_EXECUTION]: The cli.js file uses spawnSync to bootstrap the connector's execution environment. This is a standard architectural pattern for these connectors and is restricted to executing the skill's own source code or compiled artifacts using the current Node.js runtime.
[PROMPT_INJECTION]: The skill includes instructions in SKILL.md to guide the agent toward correct and safe behavior, such as disambiguating between events with similar titles and refusing unsupported bulk operations. These are beneficial safety constraints rather than malicious injection patterns.
[DATA_EXPOSURE_AND_EXFILTRATION]: As intended for a calendar tool, the skill accesses event and calendar data. Access is governed by the user's OAuth scopes and calendar roles, with logic specifically checking for sufficient permissions before attempting destructive or sharing operations.
[INDIRECT_PROMPT_INJECTION]: The skill processes external data (event titles and descriptions) that could potentially contain malicious instructions. This is an inherent risk for any tool interacting with external content; however, the skill implements structural validation via Zod schemas and provides clear guidance to the agent on how to handle such content safely.

google-calendar