The Agent Skills Directory

[SAFE]: The skill is a comprehensive documentation set for built-in platform features. No security risks were identified in the instructions or examples.
Data Exposure & Exfiltration: The skill demonstrates interaction with external services using http_request and remote_query. In all cases, it correctly utilizes the secret() function to handle credentials (e.g., bearer_token secret('scope', 'token')), preventing the leakage of sensitive keys or tokens.
Indirect Prompt Injection Surface: The documentation identifies data ingestion points such as read_files, http_request, and remote_query. While these represent an attack surface where untrusted external data could enter the agent's context, the skill itself is purely instructional and does not provide or process malicious content. (Ingestion points: read_files, http_request, remote_query in ai-functions.md; Capability inventory: execute_sql MCP tools).
Dynamic Execution: The skill provides guidance on using EXECUTE IMMEDIATE for procedural SQL. These examples follow standard administrative patterns (e.g., dynamic table truncation in sql-scripting.md) and do not involve executing untrusted code or scripts from external sources.

databricks-dbsql