The Agent Skills Directory

[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it retrieves and processes content from Notion pages and databases, which are untrusted external sources.
Ingestion points: The skill uses Notion:notion-fetch and Notion:notion-query-data-sources to pull content from the workspace into the agent's context.
Boundary markers: There are no explicit instructions or delimiters used to isolate the fetched content or warn the model to ignore any embedded instructions within those pages.
Capability inventory: The skill has write access to the workspace via Notion:notion-create-pages, Notion:notion-update-page, and Notion:notion-create-comment, which could be exploited if the agent is manipulated by injected instructions.
Sanitization: The retrieved content is not sanitized or validated before being synthesized into new meeting materials.
[EXTERNAL_DOWNLOADS]: The skill instructions include a command to add a Model Context Protocol (MCP) server from a remote URL.
Evidence: SKILL.md contains the command codex mcp add notion --url https://mcp.notion.com/mcp.
Context: The URL is the official endpoint for Notion's MCP service, which is a well-known and trusted provider for this integration.

notion-meeting-intelligence