The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from files and user-provided content that could potentially harbor malicious instructions meant to override agent behavior.\n
Ingestion points: The agent reads BUSINESS_CONTEXT.md and processes various forms of financial data provided by the user, such as P&L statements, spreadsheets, or manual figures.\n
Boundary markers: No delimiters or specific instructions are provided to help the agent distinguish between financial data and potential embedded commands, increasing the risk of instruction obedience from the data.\n
Capability inventory: The skill has file-system capabilities, specifically reading context files and writing reports to the reviews/ directory.\n
Sanitization: The skill does not implement any validation, filtering, or escaping of the ingested financial data before processing it.

financial-review