The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill facilitates the ingestion of untrusted data by providing code to extract text and tables from external PDF files.\n
Ingestion points: Code snippets in SKILL.md demonstrate reading from files such as document.pdf, input.pdf, and scanned.pdf.\n
Boundary markers: The provided examples do not include boundary markers or delimiters to help the agent distinguish between instructions and data extracted from the PDF.\n
Capability inventory: The skill provides the agent with capabilities to write files (writer.write, to_excel, c.save) and execute shell commands (qpdf, pdftk).\n
Sanitization: There is no evidence of text sanitization or filtering to prevent malicious instructions embedded in a PDF from influencing the agent's behavior.\n- [EXTERNAL_DOWNLOADS]: The documentation recommends the installation of standard Python packages (pip install pytesseract pdf2image) to handle OCR and image conversion tasks.\n- [COMMAND_EXECUTION]: The skill includes instructions for using established command-line utilities for PDF manipulation, such as pdftotext, qpdf, and pdftk.

pdf