The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill manages the research environment by installing community-standard Stata packages from the Statistical Software Components (SSC) archive at Boston College using the ssc install command.\n- [EXTERNAL_DOWNLOADS]: It references official GitHub repositories, such as anthropics/skills, as trusted sources for extending agent capabilities.\n- [COMMAND_EXECUTION]: The skill uses shell commands for its own installation (cp, ln) and provides instructions for executing external scripts and tools, including R through shell Rscript.\n- [DYNAMIC_EXECUTION]: Utilizes Stata 18's python: blocks to bridge data to specialized Python libraries like econml, causalml, and pystacked. This supports runtime execution of Python code for advanced machine learning tasks within the Stata workflow.\n- [INDIRECT_PROMPT_INJECTION]: As the skill is designed to ingest external research data in various formats (.dta, .csv, .xlsx, etc.), it possesses an attack surface for indirect prompt injection. This is mitigated by a mandatory 'Step 0' data contract that enforces strict structural and type validation on all ingested data before it is used in computations.

Full-empirical-analysis-skill-Stata