The Agent Skills Directory

[COMMAND_EXECUTION]: High-risk technical workaround in scripts/office/soffice.py. The script writes C source code to a temporary file, compiles it into a shared library using gcc, and then injects it into the LibreOffice (soffice) process via the LD_PRELOAD environment variable. This is used to shim AF_UNIX sockets in restricted virtual environments but represents a significant security surface.
[COMMAND_EXECUTION]: Frequent use of the subprocess module to execute system-level commands including gcc, soffice, pdftoppm, and git across scripts/office/soffice.py, scripts/thumbnail.py, scripts/office/pack.py, and scripts/office/validators/redlining.py.
[EXTERNAL_DOWNLOADS]: The documentation in pptxgenjs.md explicitly supports and provides examples for fetching remote images via URLs (e.g., https://example.com/image.jpg) during the slide generation process.
[PROMPT_INJECTION]: High vulnerability surface for Indirect Prompt Injection. The skill ingests untrusted data from .pptx files (unzipped and parsed as XML in unpack.py) and recommends using subagents to process this content.
Ingestion points: scripts/office/unpack.py and scripts/add_slide.py extract and manipulate raw XML from user-provided Office files.
Boundary markers: None detected in scripts; the logic interpolates extracted XML content directly into processing flows.
Capability inventory: The skill has broad capabilities including arbitrary file system access, network downloads (via pptxgenjs), and subprocess execution (soffice, gcc).
Sanitization: The skill correctly uses defusedxml to mitigate XML External Entity (XXE) attacks, but does not sanitize the natural language content within the XML that is later processed by the AI agent.

pptx