The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingest project schedule and cost data from external sources, which could contain malicious instructions designed to influence the agent's behavior during analysis.
Ingestion points: The instructions.md file specifies that the agent should accept data from user-provided CSV, Excel, and JSON files.
Boundary markers: While instructions.md contains a general instruction to "Validate inputs before processing," there are no specific prompt delimiters or instructions to ignore embedded natural language commands within the data files.
Capability inventory: The skill has filesystem permissions as defined in claw.json and utilizes the pandas library to write Excel files to the disk via the export_forecast method in SKILL.md.
Sanitization: The implementation uses Python type-hinting and basic numeric validation, but it does not sanitize string-based descriptions or item names which are subsequently interpolated into reports and exported files.

cash-flow-forecaster