The Agent Skills Directory

[SAFE]: The skill consists of educational documentation and statistical utility functions designed for experimental rigor. The provided Python code utilizes standard scientific libraries (NumPy, SciPy) for mathematical calculations such as power analysis, bootstrap intervals, and p-value corrections. The logic is transparent and follows established statistical best practices.
[COMMAND_EXECUTION]: The skill suggests using shell commands (e.g., git, grep) as diagnostic tools to audit local experimental code for statistical anti-patterns. Additionally, a utility in preregistration-and-exploratory-vs-confirmatory.md uses subprocess.run to interact with the local Git repository for verification of the study plan. These operations are standard developer workflows and are used safely within the context of ensuring experimental integrity.
[DATA_EXFILTRATION]: There are no patterns of network communication, such as use of requests, curl, or wget for external data transmission. The skill operates exclusively on local data provided to its analysis functions.
[OBFUSCATION]: No obfuscated code, hidden URLs, or suspicious encoding techniques were identified. Cryptographic hashing is used legitimately for deterministic, stable data role assignment (e.g., splitting units into train/test sets).

using-counterfactual-statistics