The Agent Skills Directory

[SAFE]: The skill operates entirely within the agent's context to process and transform financial data as described. It does not attempt to access restricted file paths, execute unauthorized shell commands, or maintain persistence on the host system.
[DATA_EXPOSURE]: The skill processes sensitive financial information including transaction amounts, account identifiers, and merchant descriptions. This is consistent with its stated purpose. There is no evidence of hardcoded credentials or data exfiltration to external domains.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted transaction descriptions which could theoretically contain malicious instructions.
Ingestion points: The incoming and existing transaction data structures defined in the input contract.
Boundary markers: The skill does not define specific delimiters for separating data from instructions, but it treats the data as structured JSON arrays.
Capability inventory: The skill does not possess capabilities to execute subprocesses, write to the filesystem, or initiate network connections, effectively neutralizing the impact of potential indirect injections.
Sanitization: The logic includes a normalization step (description_normalized) that strips metadata and standardizes formatting, which serves as a basic filter for the processed text.

transaction-deduplicator