The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits a surface for Indirect Prompt Injection. * Ingestion points: The page and query-db commands in resources/pd-notion.ts fetch untrusted content from Notion blocks and database properties. * Boundary markers: There are no boundary markers or instructions to ignore embedded commands when processing retrieved text. * Capability inventory: The skill has significant write capabilities in resources/pd-notion.ts, including create-page, update-page, append, and add-row, which could be exploited by malicious instructions found in read content. * Sanitization: The script performs minimal transformation of data (converting Notion objects to strings) without sanitizing or validating the content for executable instructions or prompt-altering patterns.

pd-notion