The Agent Skills Directory

[SAFE]: The skill incorporates robust safety features, including an 'analyzer' mode that allows users to test strategies in a simulated environment before executing real trades. It also explicitly instructs the agent to seek user confirmation for live orders.
[EXTERNAL_DOWNLOADS]: The skill depends on standard and well-regarded Python libraries for financial engineering and data processing, such as openalgo, vectorbt, TA-Lib, and duckdb. These dependencies are sourced from official repositories.
[CREDENTIALS_UNSAFE]: Secret management follows best practices by utilizing .env files and environment variables. The skill's own documentation includes rules for the agent to reject any code containing hardcoded API keys.
[DATA_EXFILTRATION]: Outbound communications via Telegram and WhatsApp are implemented as a core feature for user-defined strategy alerts. These channels are configured by the user and do not show signs of unauthorized data harvesting.
[COMMAND_EXECUTION]: Shell command execution is limited to legitimate setup operations (e.g., pip installation) and script execution for trading logic. No instances of privilege escalation or arbitrary command injection were found.

openalgo