The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted content from external sources such as social media posts, news articles, and job descriptions to enrich meeting briefs. This data processing creates a surface for indirect prompt injection. 1. Ingestion points: Data retrieved via social_post_list_live, web_search_live, and company news groups. 2. Boundary markers: There are no explicit instructions to ignore or isolate instructions embedded within external data. 3. Capability inventory: The agent can execute scripts on a remote server and access local configuration files or connected CRM data. 4. Sanitization: No specific filtering or escaping of external content is specified before processing.
[DYNAMIC_EXECUTION]: The skill's primary function involves the agent generating JavaScript code blocks which are then sent to a remote MCP server's execute tool. This dynamic script assembly is the standard method for querying the Crustdata platform but involves the runtime execution of AI-generated logic.
[DATA_EXFILTRATION]: The skill can optionally access sensitive information from a user's CRM, email, and call recordings to sharpen meeting briefs. While this access is intended for context enrichment, it involves handling sensitive business data within the agent's context.
[EXTERNAL_DOWNLOADS]: The skill connects to and retrieves data from a remote Model Context Protocol (MCP) server located at install.crustdata.com. This is the official infrastructure provided by the vendor for the skill's data retrieval and enrichment capabilities.

meeting-prep