The Agent Skills Directory

[SAFE]: No evidence of malicious behavior, data exfiltration, or prompt injection was found. The skill follows best practices by using a dedicated Python script for complex arithmetic to prevent common LLM calculation errors.
[PROMPT_INJECTION]: The skill processes external data via CSV files for NPV and IRR calculations. While this represents a surface for indirect prompt injection if user-provided files contain instructions instead of numerical data, no malicious usage or dangerous capabilities were identified in the current implementation.
Ingestion points: Reads financial data from evals/files/project_cashflows.csv and evals/files/nonconventional_cashflows.csv.
Boundary markers: No explicit delimiters or instructions are provided to the agent to ignore potential commands within the CSV files.
Capability inventory: The skill is restricted to mathematical computations and qualitative analysis; no network, file-write, or shell-execution capabilities were detected.
Sanitization: The Python calculation script implicitly validates input by attempting to cast data to floats, which provides basic structural protection.

investment-analysis