The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess calls to execute legitimate system utilities for document processing:
soffice (LibreOffice) is used in scripts/thumbnail.py and ooxml/scripts/pack.py to convert presentations to PDF or HTML for validation and thumbnail generation.
pdftoppm is used in scripts/thumbnail.py to convert PDF pages to images.
git diff is used in ooxml/scripts/validation/redlining.py to compare text changes in Word documents.
All calls use argument lists rather than shell strings, mitigating command injection risks.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user-provided .pptx files and extracts their content for analysis. While this represents a potential ingestion surface for malicious instructions hidden in document metadata or slide text, the skill employs standard safeguards:
Ingestion occurs via the markitdown package and custom XML parsing scripts (inventory.py).
Secure XML parsing is performed using the defusedxml library to prevent XML External Entity (XXE) attacks.
The risk is inherent to the skill's primary purpose and is managed by the agent's internal guardrails.
[SAFE]: The skill demonstrates safe software engineering practices:
It uses tempfile.TemporaryDirectory for intermediate processing to avoid polluting the file system.
It depends on well-known, reputable packages from official registries (PyPI and NPM).
It contains no obfuscated code, hardcoded credentials, or unauthorized network operations.

pptx