The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The skill provides instructions to install the Databricks CLI by downloading a script directly from Databricks' official GitHub repository and piping it to a shell. This is a standard installation method for this well-known service.
[EXTERNAL_DOWNLOADS]: Fetches the installation script from https://raw.githubusercontent.com/databricks/setup-cli/main/install.sh. This is an official resource from a well-known technology provider.
[COMMAND_EXECUTION]: The skill invokes several local command-line tools to perform its tasks:
databricks CLI for authentication, workspace management, and fetching SQL execution plans.
git for branch creation and committing performance fixes.
uv run to execute the local Python analyzer script and the sqlfmt formatter.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests and processes SQL files from the repository to generate reports and automated fixes.
Ingestion points: Reads .sql files using Path.rglob("*.sql") in scripts/sql_dependency_analyzer.py.
Boundary markers: Uses the sqlglot library to parse SQL structure, which provides better isolation than simple regex, though it lacks explicit "ignore instructions" delimiters for comments.
Capability inventory: The skill can write to the filesystem (SQL rewrites), execute shell commands (git commit, sqlfmt), and perform network operations via the Databricks CLI.
Sanitization: Uses structured parsing via sqlglot to extract dependencies rather than treating file content as instructions.

fabricks-sql-analyzer