skills/microsoft/hve-core/powerpoint/Gen Agent Trust Hub

powerpoint

Pass

Audited by Gen Agent Trust Hub on Sep 15, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDYNAMIC_EXECUTION
Full Analysis
  • External Downloads: The skill references a shell script to install the uv package manager from a well-known technology provider (Astral). This is a routine setup procedure for Python development environments and is considered a common practice in this ecosystem.
  • Evidence: curl -LsSf https://astral.sh/uv/install.sh | sh in SKILL.md and scripts/invoke-pptx-pipeline.sh.
  • Command Execution: The skill invokes system utilities like LibreOffice (soffice) and pdftoppm to handle document conversions. These subprocesses are managed using structured argument lists without shell interpolation, which is a defensive programming pattern that prevents command injection.
  • Evidence: subprocess.run([soffice, '--headless', ...]) in scripts/export_slides.py and scripts/export_svg.py.
  • Dynamic Execution: The presentation builder allows for the inclusion of per-slide Python scripts (content-extra.py) to perform complex drawings.
  • Evidence: spec.loader.exec_module(mod) in scripts/build_deck.py.
  • Context: This capability is gated behind a manual --allow-scripts flag and is subject to a static analysis (AST) lint that blocks dangerous modules like os and subprocess. However, the skill's own security documentation acknowledges that the lint is not a sound sandbox and that an authorized script runs with the permissions of the host process. Execution should only be granted to trusted, reviewed scripts.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 15, 2026, 02:22 PM
Security Audit — agent-trust-hub — powerpoint