powerpoint
Pass
Audited by Gen Agent Trust Hub on Sep 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill uses well-known and reputable dependencies such as markitdown, Pillow, and pptxgenjs. No suspicious or unknown packages are requested.
- [SAFE]: XML parsing is handled securely throughout the skill. Multiple scripts, including clean.py, pack.py, merge_runs.py, and simplify_redlines.py, utilize the defusedxml library to protect against XML External Entity (XXE) and other XML-based attacks.
- [INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection as it processes untrusted PowerPoint files. 1. Ingestion points: Data entering the context via markitdown extraction and XML unpacking (SKILL.md, scripts/office/unpack.py). 2. Boundary markers: Absent. 3. Capability inventory: File system writes (scripts/add_slide.py, scripts/office/pack.py) and subprocess calls for document conversion (scripts/office/soffice.py, SKILL.md). 4. Sanitization: Implementation of defusedxml for XML processing provides protection against malformed XML input.
- [SAFE]: No patterns of unauthorized command execution, data exfiltration, or persistence mechanisms were detected.
Audit Metadata