The Agent Skills Directory

[SAFE]: The skill is developed by Zapier and interacts exclusively with the official Notion API using secure, authenticated connections.
[SAFE]: Credential management is handled through connection selectors (zapier: or env:), which prevents the hardcoding of secrets. The documentation provides clear instructions to agents on how to safely verify the existence of environment variables without exposing their values in chat logs.
[COMMAND_EXECUTION]: The cli.js script facilitates the execution of the connector's tools across different environments using spawnSync. This is a standard and safe implementation for a cross-platform CLI tool.
[PROMPT_INJECTION]: The skill includes operational guidelines in SKILL.md (e.g., "Disambiguation & refusals") that help the agent maintain consistent and safe behavior, preventing it from attempting unsupported or destructive operations.
[INDIRECT_PROMPT_INJECTION]: The skill handles data from the Notion API that could contain untrusted instructions.
Ingestion points: Data is fetched from Notion in all tool scripts.
Boundary markers: The skill includes comprehensive instructional boundaries in SKILL.md for the agent to guide its interaction with Notion data.
Capability inventory: The skill has network access to the Notion API and CLI execution capabilities for running its tools.
Sanitization: Strict Zod schemas are used to validate all data entering and leaving the skill's tools.

notion