Skills
skills/nicepkg/ai-workflow/pptx/Gen Agent Trust Hub

pptx

Fail

Audited by Gen Agent Trust Hub on Feb 16, 2026

Risk Level: HIGHREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONDATA_EXFILTRATION
Full Analysis
  • COMMAND_EXECUTION (HIGH): Zip Slip vulnerability in ooxml/scripts/unpack.py. The script uses zipfile.ZipFile.extractall() on user-provided documents without path validation, allowing an attacker to write files to arbitrary locations via path traversal.
  • Evidence: ooxml/scripts/unpack.py line 17
  • REMOTE_CODE_EXECUTION (HIGH): Indirect Prompt Injection surface. The skill ingests untrusted OOXML data and possesses write/execute capabilities. A malicious document could exploit the Zip Slip vulnerability to overwrite skill scripts or system files to achieve code execution.
  • Ingestion points: ooxml/scripts/unpack.py extracts external ZIP/OOXML content.
  • Boundary markers: Absent.
  • Capability inventory: File writes (extractall), XML manipulation, and subprocess execution (soffice).
  • Sanitization: Absent for ZIP paths; inconsistent for XML (uses lxml instead of defusedxml in docx.py).
  • DATA_EXFILTRATION (MEDIUM): XXE (XML External Entity) vulnerability in ooxml/scripts/validation/docx.py. The use of lxml.etree.parse() without disabling entity resolution allows an attacker to read local files or trigger SSRF requests via malicious XML.
  • Evidence: ooxml/scripts/validation/docx.py lines 92, 127
  • COMMAND_EXECUTION (LOW): Subprocess call to soffice in ooxml/scripts/pack.py. While the call structure is generally safe, it introduces a dependency on the security and behavior of an external office suite when processing untrusted files.
  • Evidence: ooxml/scripts/pack.py line 106
Recommendations
  • AI detected serious security threats
Audit Metadata
Risk Level
HIGH
Analyzed
Feb 16, 2026, 04:25 AM