The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external data files (CSV, Excel) provided by the user, which presents an attack surface for indirect prompt injection.
Ingestion points: The extract_spend_data function in SKILL.md reads from user-provided file paths or dataframes using pd.read_csv and pd.read_excel.
Boundary markers: The instructions do not define delimiters or specific warnings to ignore instructions that might be embedded within the spend data files.
Capability inventory: The skill's scripts utilize pandas, scikit-learn, and matplotlib for analysis and visualization. No file writing, network operations, or shell command execution capabilities are present in the provided code.
Sanitization: The clean_spend_data function performs standardization of supplier names and date/amount parsing, but does not filter for potential prompt injection payloads in text fields like descriptions.

spend-analysis