econometrics-skills
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill references external repositories and package managers to facilitate the setup of econometric analysis environments. These references involve well-known services and academic research repositories.\n
- Fetches the
did_multiplegtStata package from a GitHub repository hosted by academic researchers.\n - Recommends the installation of the
econmlpackage, an official Microsoft Research project, viapipfor advanced causal inference tasks.\n - Directs users to clone the
python-causality-handbookrepository from GitHub for access to interactive tutorials.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided data and research parameters, creating a surface for potential indirect prompt injection.\n - Ingestion points: User-provided datasets and variable names are ingested via agent tools and processed using templates in guide files such as
causal-inference-guide/SKILL.md.\n - Boundary markers: The provided code templates do not implement specific boundary markers or instructions to isolate or ignore potentially malicious instructions embedded in data.\n
- Capability inventory: The skill provides templates that utilize file writing (e.g.,
graph export,esttab using) and statistical computation tools across the package guide files.\n - Sanitization: The logic for generating model formulas (e.g., in
causal-inference-guide/SKILL.md) involves direct interpolation of variable names, which lacks validation or escaping mechanisms for external inputs.
Audit Metadata