The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses a vulnerability surface for indirect prompt injection as it is designed to process and render untrusted content from external HTML and PPTX files using powerful execution environments.
Ingestion points: Untrusted data enters the agent context through scripts/html2pptx.js (reading HTML files) and through Python scripts such as inventory.py and replace.py (processing PPTX files).
Boundary markers: The skill does not implement explicit boundary markers or "ignore embedded instructions" warnings when processing the text content of slides or HTML templates.
Capability inventory: The skill has access to a headless browser (playwright), file system write operations, and several command-line tools executed via subprocesses including soffice (LibreOffice), pdftoppm, and git (in pack.py, thumbnail.py, and redlining.py).
Sanitization: While the skill correctly uses defusedxml to mitigate XML-based attacks during parsing, it does not fully sanitize interpolated text content against nested natural language instructions that might influence the agent's behavior during analysis.

pptx