tts-voiceover

Pass

Audited by Gen Agent Trust Hub on Sep 16, 2026

Risk Level: SAFEEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
  • External Tool Installation Recommendation: The shell and PowerShell wrapper scripts include an error message suggesting the installation of uv (a Python package manager) via a curl | sh command.\n
  • Evidence: Found in scripts/generate-voiceover.sh, scripts/embed-audio.sh, and scripts/Modules/TtsVoiceoverHelpers.psm1 pointing to https://astral.sh/uv/install.sh.\n
  • Context: While executing remote scripts involves risk, astral.sh is the official provider for the well-known uv tool. The scripts do not auto-execute this command; it is provided as a setup instruction for the user to manually perform if the dependency is missing.\n- Data Egress to Azure Services: The skill transmits speaker-notes text to Azure Speech endpoints for audio synthesis.\n
  • Evidence: Logic implemented in scripts/generate_voiceover.py using the azure-cognitiveservices-speech SDK.\n
  • Context: This is the core functionality of the skill. The data egress is transparently acknowledged in the SECURITY.md file, which advises users to avoid sending regulated or confidential narration to ensure compliance with data sovereignty requirements.\n- Input Sanitization and Injection Mitigation: The skill processes YAML data and generates SSML documents.\n
  • Evidence: scripts/generate_voiceover.py uses yaml.safe_load for parsing and xml.sax.saxutils for escaping speaker notes and acronym aliases.\n
  • Context: These practices prevent common injection vulnerabilities, such as SSML injection or arbitrary code execution via unsafe YAML deserialization, when processing user-supplied slide content.\n- Hardened Document Parsing: The skill parses PowerPoint (OOXML) files and internal XML templates using potentially risky parsers.\n
  • Evidence: scripts/embed_audio.py explicitly configures lxml.etree.XMLParser(resolve_entities=False, no_network=True).\n
  • Context: This hardening prevents XML External Entity (XXE) vulnerabilities, ensuring that malicious inputs cannot trigger unauthorized file access or network requests during the audio embedding process.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 16, 2026, 06:56 AM
Security Audit — agent-trust-hub — tts-voiceover