The Agent Skills Directory

[SAFE]: The skill consists entirely of documentation, architectural patterns, and code examples for developers using MLflow 3 and Databricks. It does not contain any executable scripts, automated shell commands, or obfuscated code.
[PROMPT_INJECTION]: The static analysis hits for prompt injection in references/patterns-datasets.md and references/user-journeys.md are false positives. The phrases "Ignore previous instructions" and "What is your system prompt?" are explicitly categorized as examples of "Adversarial Cases" that developers should include in their evaluation datasets to test the robustness of their agents. These are data samples for testing, not instructions to the AI agent.
[CREDENTIALS_UNSAFE]: The skill demonstrates secure credential handling by using placeholders such as <YOUR_API_TOKEN> and <SQL_WAREHOUSE_ID> in code examples. It correctly advises users to manage sensitive information via environment variables rather than hardcoding them into source code.
[EXTERNAL_DOWNLOADS]: The documentation references standard installations of established libraries like mlflow and openai from official package registries. All external URLs point to official Databricks documentation, the MLflow GitHub organization, or Databricks-owned domains, which are trusted sources for this skill authored by Databricks.

databricks-mlflow-evaluation