The Agent Skills Directory

[SAFE]: The skill provides instructional content and setup scripts for Databricks Lakebase using official tools. No malicious intent or suspicious patterns were detected.
[COMMAND_EXECUTION]: The skill includes instructions for running databricks CLI commands and uv to deploy resources and manage environment configurations. These are standard operations for Databricks application development.
[DYNAMIC_EXECUTION]: Python initialization scripts are executed via python -c and shell heredocs. These scripts are used for one-time table setup and permission granting during the configuration phase, which is standard practice for this type of infrastructure-as-code setup.
[DATA_EXPOSURE]: The skill references .env files and environment variables for configuration. All sensitive values (like instance names and client IDs) are provided as placeholders, following security best practices for documentation.
[PROMPT_INJECTION]: The skill facilitates the creation of a persistent memory store (LangGraph Store), which introduces an indirect prompt injection surface by allowing agents to ingest and act upon historically stored user data. This is a characteristic of memory-enabled agents rather than a vulnerability in the skill itself.
Ingestion points: AsyncDatabricksStore used in SKILL.md to persist and retrieve agent memories.
Boundary markers: No specific boundary markers or 'ignore' instructions are defined in the setup templates.
Capability inventory: Read and write access to Lakebase PostgreSQL tables (store, store_vectors).
Sanitization: Sanitization and validation of retrieved memory are expected to be implemented within the agent logic that utilizes this store.

lakebase-setup