The Agent Skills Directory

COMMAND_EXECUTION (HIGH): Path Traversal (ZipSlip) vulnerability detected in multiple locations.\n
Evidence (ooxml/scripts/unpack.py): The script uses zipfile.ZipFile(input_file).extractall(output_path) on user-provided Office documents.\n
Evidence (ooxml/scripts/validation/docx.py): The function count_paragraphs_in_original also uses zip_ref.extractall(temp_dir).\n
Risk: Malicious ZIP archives (which Office files are) can contain filenames with path traversal sequences (e.g., ../../) that allow an attacker to overwrite arbitrary files on the system during extraction, potentially leading to remote code execution or system compromise.\n- DATA_EXFILTRATION (MEDIUM): Potential XML External Entity (XXE) vulnerability in ooxml/scripts/validation/docx.py.\n
Evidence: The code uses lxml.etree.parse(str(xml_file)) to process XML components. Unlike the defusedxml library used elsewhere in the skill, lxml.etree by default may resolve external entities.\n
Risk: An attacker providing a crafted document could use malicious XML to read local system files or perform Server-Side Request Forgery (SSRF) when the agent validates the document.\n- COMMAND_EXECUTION (MEDIUM): Subprocess execution of an external office suite in ooxml/scripts/pack.py.\n
Evidence: The validate_document function executes soffice (LibreOffice) via subprocess.run to perform conversion checks.\n
Risk: While the script uses a list of arguments to mitigate shell injection, processing untrusted complex file formats with a large office suite exposes the agent to vulnerabilities within the suite's own parsers.

pptx