notion-meeting-intelligence
Pass
Audited by Gen Agent Trust Hub on Sep 18, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill demonstrates an attack surface for indirect prompt injection by design. It retrieves data from external sources (Notion pages and databases) and uses that information to generate new documents.
- Ingestion points: The skill uses
Notion:notion-fetchto read page content (SKILL.md, Step 3) andNotion:notion-query-data-sourcesto retrieve database records (evaluations/status-meeting-prep.json, Step 4). - Boundary markers: The instructions lack specific guidance for the agent to use delimiters or to ignore instructions embedded within the fetched Notion content, increasing the risk of the agent following instructions found in those pages.
- Capability inventory: The skill utilizes tools to create new Notion pages (
Notion:notion-create-pages) and add comments (Notion:notion-create-comment), which could be used to propagate malicious content. - Sanitization: There are no explicit instructions for the agent to sanitize or validate the content retrieved from Notion before incorporating it into the meeting materials.
Audit Metadata