The Agent Skills Directory

[SAFE]: The skill is a well-documented analysis module for causal inference. It adheres to academic best practices by requiring identifying assumptions to be stated and refuted alongside estimation. All scripts and references are functional and benign.
[EXTERNAL_DOWNLOADS]: The skill depends on standard, well-known Python packages from the data science ecosystem, including statsmodels, linearmodels, pyfixest, dowhy, econml, and rdrobust. These are reputable libraries used for research and econometrics.
[COMMAND_EXECUTION]: The skill utilizes the Bash(python:*) tool to run statistical models. This is the intended and necessary functionality for an analysis module and does not involve unauthorized system access or network activity.
[DYNAMIC_EXECUTION]: The skill provides the agent with static Python code templates located in references/estimators.md. These templates are populated with local dataset variables at runtime to perform calculations. No dynamic loading of untrusted code or arbitrary remote execution was observed.
[INDIRECT_PROMPT_INJECTION]: The skill ingests user-provided datasets which represents a potential attack surface for indirect prompt injection.
Ingestion points: External data enters the agent context through the data parameter in library calls (e.g., smf.ols or pf.feols) as described in references/estimators.md.
Boundary markers: Absent; data content is processed directly by the statistical libraries.
Capability inventory: Local Python execution via Bash(python:*) as specified in SKILL.md.
Sanitization: Absent; the skill relies on the robustness of the statistical libraries to handle data inputs correctly without executing embedded strings as instructions.

alterlab-causal-inference