The Agent Skills Directory

[DATA_EXPOSURE]: The skill manages authentication using OAuth 2.0 tokens. It supports both Zapier-managed connections (zapier:connection-id) and direct token resolution via environment variables (env:ENV_VAR). The documentation (SKILL.md and references/use-without-zapier.md) explicitly instructs the agent not to dump environment variables to avoid leaking credentials in chat transcripts.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it retrieves data (summaries, descriptions) from external calendar events. A malicious event description could contain instructions designed to influence agent behavior. However, this is inherent to the tool's primary purpose of reading calendar data, and the skill provides clear guidance on resolving ambiguities and refusing unsupported bulk operations to mitigate accidental mis-execution.
[COMMAND_EXECUTION]: The cli.js file uses spawnSync to execute the connector's own TypeScript source or compiled distribution. This execution is restricted to the current Node/Bun runtime and the skill's internal files, serving as a dispatcher rather than an arbitrary command execution vector.
[EXTERNAL_DOWNLOADS]: All network operations are directed at official Google APIs (https://www.googleapis.com/calendar/v3/) and the author's own infrastructure. These are well-known, trusted services for a calendar management skill.

google-calendar