The Agent Skills Directory

[COMMAND_EXECUTION]: Several scripts, including batch.py, bootstrap.py, and collect_results.py, use the subprocess.run method to interact with the local pandaai-cli tool. These operations are restricted to executing the specific, documented subcommands required for factor creation, result retrieval, and balance checking. The command structures are static, using user-provided identifiers only as positional arguments to the CLI tool.
[EXTERNAL_DOWNLOADS]: The bootstrap.py script performs a read-only metadata check by running git fetch against the skill's official GitHub repository. This is used solely to inform the user if a newer version of the skill is available and does not involve the automatic download or execution of remote code.
[CREDENTIALS_SAFE]: The skill instructions and scripts prioritize credential security. The agent is explicitly directed to have the user enter sensitive login information (phone and password) in their own host terminal. Scripts check for the presence of authentication tokens in the local ~/.pandaai/config.yaml file to verify login status without ever reading, logging, or exfiltrating the credentials themselves.
[DYNAMIC_EXECUTION]: The build_composites.py script includes functionality to generate Python source code representing composite financial factors. This generation process uses deterministic templates to combine user-provided factor components and is intended for user review before being uploaded to the PandaAI platform. No untrusted input is executed locally during this process.

pandaai-factor-online