The Agent Skills Directory

[SECURITY_BEST_PRACTICE]: The skill correctly demonstrates path sanitization using os.path.basename() when extracting embedded attachments from PDFs, which prevents path traversal attacks from malicious files.
[COMMAND_EXECUTION]: The skill utilizes standard, well-known command-line utilities for PDF processing, including pdfinfo, pdftotext, pdfimages, pdfdetach, pdffonts, pdftoppm, and pdftk. These tools are industry standards for document analysis.
[INDIRECT_PROMPT_INJECTION]: As a tool designed to ingest untrusted data (PDF content), the skill inherently creates an attack surface for indirect prompt injection. While the skill does not explicitly provide prompt delimiters for the extracted text, it focuses on the technical extraction methods rather than the agent's internal prompt handling.
[DATA_EXPOSURE]: The skill instructions involve reading local PDF files and writing extracted components (images, attachments, text) to the local /tmp/ directory. No network exfiltration or unauthorized access patterns were detected.

pdf-reading