The Agent Skills Directory

Indirect Prompt Injection: The skill is designed to fetch and analyze content from external sources, specifically Notion pages and databases, to generate meeting agendas and pre-reads. This capability introduces a potential surface for indirect prompt injection, where malicious instructions hidden in a Notion document could attempt to override the agent's behavior during synthesis.
Ingestion points: Data enters the agent's context through Notion:search, Notion:fetch, and Notion:notion-query-data-sources calls as defined in SKILL.md and the examples/ directory.
Boundary markers: The skill instructions do not currently implement specific delimiters or instructions to ignore potential commands embedded within the retrieved Notion content, which could lead the agent to inadvertently follow instructions found in the data.
Capability inventory: The agent has the ability to create new Notion pages (Notion:notion-create-pages), update existing content (Notion:notion-update-page), and leave comments (Notion:notion-create-comment), providing a mechanism for an injection to persist or affect the workspace.
Sanitization: There is no explicit evidence of content sanitization or validation logic applied to the data fetched from Notion before it is processed by the model.

notion-meeting-intelligence