The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill's primary workflow for creating presentations from scratch involves the automated generation of multiple JavaScript modules (e.g., slide-01.js) and a main execution script (compile.js). These files are then executed at runtime using node compile.js via the shell, which constitutes a dynamic code execution pattern.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process untrusted PowerPoint files using the markitdown tool. The extracted text and structural data are then used to inform the agent's planning and generation of new presentation content. This creates a vulnerability surface where instructions embedded within a source presentation could attempt to manipulate the agent's behavior during the generation phase.
Ingestion points: PowerPoint files processed via markitdown in SKILL.md and references/editing.md.
Boundary markers: No explicit instruction delimiters or safety warnings are specified for the interpolation of extracted content into the subagent prompts.
Capability inventory: File system read/write, XML manipulation using zipfile and defusedxml, and shell execution of Python and Node.js scripts.
Sanitization: The skill recommends using defusedxml to mitigate XML-based attacks, but does not specify sanitization or validation for the natural language content extracted from presentation slides.
[COMMAND_EXECUTION]: The skill frequently utilizes shell commands to perform its core functions, including python -m markitdown for text extraction and node compile.js for presentation assembly.

pptx-generator