The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external datasets (Pandas DataFrames) and strategy configurations in evaluate_exit_factor and VectorBacktester. While this represents a potential surface for indirect prompt injection if processing untrusted user data, the implementation lacks risky automated actions following the data ingestion. Evidence found in exit_analysis.py and vector.py.
[DYNAMIC_EXECUTION]: The evaluate_exit_factor function in exit_analysis.py utilizes dependency injection by accepting a backtester_cls and exit_filter callables. This is a standard architectural pattern for modular research frameworks and does not involve unsafe string evaluation or execution of untrusted remote code.
[COMMAND_EXECUTION]: The skill performs temporary redirection of sys.stdout to a buffer (io.StringIO) in exit_analysis.py within the _run_backtest_silent function to suppress console output during strategy execution. This is a benign utility function for cleaner logs and does not execute shell commands or exfiltrate data.

backtest