The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests external data from user-provided CSV files, which serves as a vector for indirect prompt injection attacks if the data contains malicious instructions.
Ingestion points: Transaction descriptions and customer email fields from the Stripe CSV are imported into the agent's environment as described in the Workflow and Column Mapping sections of SKILL.md.
Boundary markers: The skill lacks explicit delimiters or instructions to the agent to disregard potential commands embedded within the transaction data.
Capability inventory: The skill utilizes the categorize and transaction_search tools, which may process the untrusted strings to make decisions or modifications to the accounting database.
Sanitization: The instructions do not specify any validation, filtering, or escaping of the CSV cell content before it is handled by the agent's tools.

stripe-import