The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from the Notion API, which acts as a potential vector for malicious instructions if the source workspace is compromised. \n- Ingestion points: The paginateDatabase generator in references/implementation.md retrieves content via notion.databases.query. \n- Boundary markers: The implementation does not include specific delimiters or prompt-isolation instructions to protect against instructions embedded in API results. \n- Capability inventory: The skill utilizes file system operations (fs.writeFileSync) and is allowed command execution tools (Bash). \n- Sanitization: No data sanitization or structural validation is performed on the API responses before they are processed by the worker. \n- [EXTERNAL_DOWNLOADS]: The skill references and requires official, well-known dependencies for its operation, specifically @notionhq/client and notion-client, as well as the standard p-queue library. \n- [COMMAND_EXECUTION]: A diagnostic throughput benchmark in references/examples.md uses the curl utility to perform sequential network requests to the official Notion API for latency measurement.

notion-load-scale