The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The instructions recommend using sudo to install system-level language packages, which involves executing commands with administrative privileges.\n
Evidence: sudo apt install tesseract-ocr-deu tesseract-ocr-jpn tesseract-ocr-fra in SKILL.md.\n- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process text from external, untrusted sources such as scanned PDFs and images via OCR, which creates an attack surface for indirect prompt injection.\n
Ingestion points: xberg extract is applied to files like scan.pdf, menu.jpg, and other user-provided documents.\n
Boundary markers: Absent; the instructions do not define delimiters or warnings to treat the extracted text as untrusted data.\n
Capability inventory: The skill utilizes the xberg CLI tool and system package managers (apt, brew).\n
Sanitization: Absent; there are no steps mentioned for sanitizing or validating the output produced by the OCR backends before it is consumed by the agent.\n- [COMMAND_EXECUTION]: The skill involves the execution of the xberg CLI tool and OS package managers to perform its core functions and setup.\n
Evidence: Example commands including xberg extract, xberg batch, brew install, and apt install are central to the skill's usage.

extracting-with-ocr