The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection because it ingests untrusted data and possesses significant system capabilities.
Ingestion points: Reads data from external files using use, import excel, and import delimited commands (e.g., in references/01-data-cleaning.md).
Boundary markers: The skill does not implement delimiters or specific instructions to the agent to ignore embedded instructions within the processed data files.
Capability inventory: The skill utilizes shell commands (for R scripts), python: blocks (for ML libraries), and extensive file operations (save, export, file write) across multiple reference files and the main playbook.
Sanitization: There is no evidence of sanitization or validation of the content of imported data files before processing.
[REMOTE_CODE_EXECUTION]: The skill uses the ssc install command to download and install community-developed packages from the SSC (Boston College) Stata repository (e.g., in README.md and SKILL.md).
[COMMAND_EXECUTION]: The skill executes external R scripts using the shell Rscript command for policy learning tasks (e.g., in references/07-further-analysis.md).
[DYNAMIC_EXECUTION]: The skill utilizes Stata 18's python: blocks to execute Python code (using libraries like econml and causal-learn) within the Stata process environment (e.g., in references/07-further-analysis.md and references/05-modeling.md).
[EXTERNAL_DOWNLOADS]: The skill fetches datasets from remote URLs using the use command (e.g., fetching crime data from stats.idre.ucla.edu in references/01-data-cleaning.md).

Full-empirical-analysis-skill-Stata