The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes external Excel files containing financial records. Malicious instructions embedded within account names or transaction descriptions could be reflected in the audit report, potentially influencing the agent's interpretation of the results.
Ingestion points: The script scripts/reconcile.py ingests data from multiple user-supplied Excel files through arguments like --balance-sheet and --account-changes.
Boundary markers: Output reports (console and Excel) do not employ delimiters or specific warnings to isolate data from agent instructions.
Capability inventory: The skill utilizes file system read/write operations for processing and reporting but lacks network access or shell execution capabilities beyond its own script.
Sanitization: The script performs type conversion for numeric values but lacks content sanitization for string fields like account names or items.

finance-reconcile