The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill utilizes several external libraries from both the Python and Node.js ecosystems, including markitdown[pptx] (an official Microsoft tool), pptxgenjs, react-icons, react, react-dom, and sharp. In accordance with safety rules, these established dependencies from reputable sources are documented as part of the skill's functional requirements.
[DYNAMIC_EXECUTION]: The workflow requires the generation of discrete JavaScript files for each slide, which are subsequently loaded and executed by a compilation script using the require() function. This dynamic loading of locally generated scripts is a core operational mechanism to facilitate complex, automated presentation building.
[INDIRECT_PROMPT_INJECTION]: The skill ingests unverified data from external PowerPoint files via text extraction. This content is then used to structure new slides and inform the generation of executable code, creating a potential vector for indirect injection attacks.
Ingestion points: Text extracted from user-supplied presentations using markitdown as described in SKILL.md and references/editing.md.
Boundary markers: The provided instructions lack specific delimiters or instructions to ignore embedded commands when processing extracted metadata or slide text.
Capability inventory: The skill possesses the capability to write and execute JavaScript code, perform significant file system operations, and modify XML structures.
Sanitization: While the technical documentation recommends using defusedxml to mitigate standard XML vulnerabilities during the editing process, there are no defined procedures for sanitizing or isolating natural language content extracted from the input presentations.

pptx-generator