The Agent Skills Directory

[SAFE]: The skill comprises high-quality instructional content for data engineering professionals. All code examples for DuckDB, Polars, and SQL follow industry-standard security practices, such as idempotency and logical-time parameterization. No unauthorized network requests, credential harvesting, or obfuscated code were found during the analysis.\n- [PROMPT_INJECTION]: The skill features an 'AUDIT mode' that instructs the agent to read and evaluate external repository content, including configurations, models, and SQL code. This ingestion of potentially untrusted data represents an indirect prompt injection surface, which is a necessary and intended part of the auditing functionality. Ingestion points: SKILL.md (lines 67-73) directs the agent to inventory and read orchestrator configurations, scheduler configurations, transformation tools, and DAG/model files. Boundary markers: The skill does not specify particular delimiters or 'ignore' instructions for the ingested code. Capability inventory: The agent's capabilities are focused on analysis and reporting; it does not request high-privilege system access or network permissions. Sanitization: The instructions do not define methods for sanitizing the audited data before processing.

sota-data-engineering