The Agent Skills Directory

[COMMAND_EXECUTION]: The skill scaffolds a Makefile that uses python3 -m http.server for local development and git commands for deployment to GitHub Pages. These are standard developer operations and are controlled by the user via the make utility.
[EXTERNAL_DOWNLOADS]: In Slidev mode, the skill executes npm install to fetch the @slidev/cli and @slidev/theme-default packages. These are well-known, legitimate packages from the official NPM registry.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it ingests data from project files to generate slide content. 1. Ingestion points: The agent reads README.md, docs, or solution.md files to derive content for slides.md or presentation.html. 2. Boundary markers: The skill uses markdown and HTML structure as boundaries but does not include explicit warnings against embedded instructions in the source data. 3. Capability inventory: The skill can write files and initiate command-line operations (npm, git, python) through the provided Makefile. 4. Sanitization: No specific input sanitization is performed on the ingested documentation content.
[DYNAMIC_EXECUTION]: The provided HTML slide templates in the references/ directory include JavaScript for in-browser editing and file saving via Blob-based downloads. This execution is limited to the user's browser environment and does not pose a risk to the agent's host system.

slides