skills/azeem-2/hackthonii/pptx/Gen Agent Trust Hub

pptx

Warn

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: MEDIUMPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: Deceptive metadata is present in the LICENSE.txt file, which claims the skill is copyrighted by Anthropic, PBC, despite the author being 'Azeem-2'. This potentially misleads users regarding the skill's trust level. The skill also facilitates indirect prompt injection by processing untrusted HTML (scripts/html2pptx.js) and PPTX data (scripts/inventory.py) without sanitization or boundary markers. Capability inventory includes shell execution (soffice, git) and file writes.
  • [COMMAND_EXECUTION]: The skill invokes system binaries like soffice, pdftoppm, and git for document processing via subprocess.run. Furthermore, ooxml/scripts/unpack.py uses zipfile.extractall() without member validation, creating a path traversal vulnerability (Zip Slip) if the agent processes a malicious presentation file.
  • [EXTERNAL_DOWNLOADS]: Documentation instructs users to install standard dependencies including markitdown, playwright, and pptxgenjs from official package registries.
  • [SAFE]: Secure XML parsing is implemented using the defusedxml library in pack.py and unpack.py to mitigate XXE risks.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 15, 2026, 11:04 AM