The Agent Skills Directory

[SAFE]: The skill exhibits a high security posture by providing explicit instructions to the AI agent to treat all extracted text as data, not instructions. This directly addresses the risk of indirect prompt injection from document content.\n- [COMMAND_EXECUTION]: The export script is used for structured data export and validation. It employs defensive programming techniques such as strict regex validation for identifiers, SHA-256 checksum verification, and prevention of file overwrites.\n- [INDIRECT_PROMPT_INJECTION]: The skill implements a specific mitigation for CSV injection (Formula Injection) by automatically escaping potentially malicious characters in the export script. This prevents the output from being used to execute arbitrary formulas in spreadsheet applications.\n- [DATA_EXFILTRATION]: The skill explicitly prohibits uploading documents to external services or OCR endpoints, ensuring that sensitive data remains within the local environment.

document-extract