The Agent Skills Directory

Indirect Prompt Injection Surface: The skill processes external data retrieved from Notion pages to generate meeting agendas and pre-reads. This creates a potential vulnerability where malicious instructions embedded in a Notion document could influence the agent's behavior during the synthesis process.
Ingestion Points: The skill uses Notion:notion-fetch (documented in SKILL.md and examples/) to ingest content from arbitrary Notion pages in the user's workspace.
Capability Inventory: The agent has the ability to create and update Notion pages (Notion:notion-create-pages, Notion:notion-update-page) and post comments (Notion:notion-create-comment), meaning injected instructions could potentially lead to unauthorized content modification or data leakage via comments.
Boundary Markers: The templates provided in the reference/ directory do not currently include explicit instructions for the model to ignore any natural language commands found within the fetched Notion content.
Sanitization: There is no evidence of filtering or sanitization of the retrieved content before it is interpolated into the prompts for summary and document generation.

notion-meeting-intelligence