The Agent Skills Directory

[COMMAND_EXECUTION]: The skill includes local Python scripts (backtest.py and validate_report.py) designed to perform statistical analysis and structural validation of research reports.
[INDIRECT_PROMPT_INJECTION]: The skill processes external CSV data provided by the user, which is a potential attack surface for indirect prompt injection.
Ingestion points: User-supplied price and signal CSV files are loaded via the load_prices and load_signal functions in scripts/backtest.py.
Boundary markers: No explicit delimiters are used in the generated markdown reports, but the agent is guided by structured templates in the references/ directory.
Capability inventory: The skill performs local file read/write operations within the workspace; it has no network access or high-privilege execution capabilities.
Sanitization: scripts/backtest.py uses pd.to_numeric with errors='coerce' and pd.to_datetime, which effectively neutralizes text-based injection payloads by converting non-numeric content into NaN values or raising errors if required columns are missing.
[DATA_EXFILTRATION]: The skill's documentation and source code confirm an offline-only design. Analysis of the Python scripts shows no use of networking libraries such as requests, urllib, or socket-level operations.

global-macro-trend-strategy