The Agent Skills Directory

[SAFE]: The skill instructions and associated code are focused on the stated purpose of financial calculations and contain no evidence of malicious intent.
[PROMPT_INJECTION]: The skill defines a surface for potential indirect prompt injection through its data ingestion points.
Ingestion points: Portfolio positions and cost basis data processed in scripts/tax_efficiency.py.
Boundary markers: None present in the instructions.
Capability inventory: The skill metadata allows use of Bash, Read, Write, and Edit tools.
Sanitization: The script explicitly casts input arrays to np.float64 within the TaxLossHarvesting class, ensuring only numeric data is processed and preventing any text-based instructions from influencing agent behavior.
[SAFE]: The script uses the trusted numpy library and performs no network operations, remote code execution, or sensitive file accesses.

tax-efficiency