The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests raw SQL strings which could serve as a vector for indirect instructions if the input is attacker-controlled.
Ingestion points: The --sql argument passed to uv run lkr parse sql-lookml-compare in SKILL.md.
Boundary markers: The instructions do not define specific delimiters or "ignore" directives to isolate the SQL content from the agent's logic.
Capability inventory: The skill has the ability to create and modify files in the repository using lkr tools lookml create and execute Python scripts via lkr code-mode sandbox.
Sanitization: There is no evidence of input validation or sanitization for the SQL query within the skill's instructions.
[COMMAND_EXECUTION]: The skill executes vendor-specific CLI tools lkr and lkr-dev-cli using uv and uvx for parsing SQL and managing LookML projects. These tools are associated with the skill author 'lkrdev'.
[DYNAMIC_EXECUTION]: The skill uses lkr code-mode sandbox to execute a local Python script scripts/generate_bronze_view.py at runtime. The script dynamically handles variables like project_id, schema_name, and table_name injected by the platform environment.

generate-view-from-table