The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill clones source code from an external repository at https://github.com/howdymary/autopredict.git during the setup process.\n- [REMOTE_CODE_EXECUTION]: The setup.sh script executes an editable pip installation on the cloned content, which triggers any build-time code execution defined in the repository.\n- [COMMAND_EXECUTION]: The skill uses shell scripts to wrap Python CLI operations for market scanning and backtesting, executing these via system-level shell commands.\n- [DYNAMIC_EXECUTION]: In scripts/tune_params.sh, a Python script is embedded and executed dynamically to facilitate parameter grid searching and configuration generation.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted market data and user datasets, posing a risk if these sources contain malicious instructions.\n
Ingestion points: External data enters via predict.py and the backtest CLI sub-command.\n
Boundary markers: The provided shell wrappers do not use boundary markers or explicit isolation for the data being processed.\n
Capability inventory: The skill has the capability to execute shell commands, access the network, and write to the local file system.\n
Sanitization: The wrapper scripts do not perform sanitization on the data prior to execution or analysis.

autopredict