The Agent Skills Directory

[DATA_EXFILTRATION]: The skill communicates exclusively with the official Notion API at api.notion.com. It provides extensive security documentation, advising users to store credentials in environment variables and warning against shell commands that could leak secrets into transcript logs.
[DYNAMIC_EXECUTION]: The entry script cli.js uses node:child_process to execute the connector's internal logic. This is a standard architectural pattern for dispatching commands in this connector framework and does not involve executing untrusted code or user-provided scripts.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it retrieves content from external Notion pages.
Ingestion points: Data is ingested through scripts/getPageAsMarkdown.ts, scripts/getBlockChildren.ts, and scripts/queryDataSource.ts.
Boundary markers: The tool returns data within a structured { data, meta } envelope, though it does not insert specific natural language boundary markers into the raw content.
Capability inventory: The skill possesses full CRUD capabilities for Notion workspaces across 24 scripts, including page creation and block modification.
Sanitization: All inputs and outputs are validated using strict Zod schemas. The scripts/copyPage.ts tool specifically strips non-writable metadata fields from block objects during cross-workspace transfers.
[REMOTE_CODE_EXECUTION]: The installation instructions utilize npx to fetch the official @zapier/notion-connector package. This is a trusted distribution method for the vendor and does not introduce unverifiable dependencies.

notion