The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external data files and user-provided variable names to generate regression specifications, which presents a surface for potential injection if the data source is untrusted.
Ingestion points: User-provided variable names (regressors/controls) and Stata data files (.dta) defined in SKILL.md.
Boundary markers: None specified; instructions rely on user input to define the analysis scope.
Capability inventory: Generates commands for file reading (use), statistical modeling (regress, reghdfe), and file writing/exporting (esttab).
Sanitization: No explicit sanitization of input strings or variable names is performed by the skill instructions.
[SAFE]: The skill utilizes the standard Stata ssc install mechanism to manage dependencies for estout and reghdfe, which are established, well-known tools in the econometrics community.
[SAFE]: All external references point to official documentation (stata.com), well-known academic archives (repec.sowi.unibe.ch), or reputable community development repositories on GitHub.

stata-regression