The Agent Skills Directory

[DYNAMIC_EXECUTION]: The run_backtest.py script uses importlib.import_module to load user-configured strategy classes from local files, which is a required functional component for customizable backtesting simulations.
[COMMAND_EXECUTION]: Multiple scripts, including financial_cli.py and fetch_data.py, utilize subprocess.run to call platform-specific authentication tools and local CLI helpers for data retrieval.
[EXTERNAL_DOWNLOADS]: The setup.sh script installs the nautilus_trader library and requests package from official registries. These are well-known libraries used for the skill's primary simulation functionality.
[INDIRECT_PROMPT_INJECTION]:
Ingestion points: The skill ingests untrusted news headlines and long-form research articles through the research/news and research/search endpoints.
Boundary markers: The SKILL.md instructions explicitly direct the agent to manually judge the tone of news headlines, maintaining a separation between data and instructions.
Capability inventory: The skill maintains capabilities for local command execution via subprocess and database interaction via a read-only SQL endpoint.
Sanitization: The financial_cli.py script enforces a strict _validate_read_only_sql check that blocks mutating keywords like DELETE, DROP, and UPDATE, effectively preventing accidental or malicious database modifications.

stock-analysis