The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill documents features that allow for the ingestion of untrusted data while providing the agent with file-writing capabilities.
Ingestion points: The skill supports importing external code snippets (references/code-import-snippet.md) and embedding content via iframes (references/core-layouts.md).
Boundary markers: There are no specific instructions or delimiters mentioned to prevent the agent from executing instructions embedded within these external sources.
Capability inventory: The skill provides a built-in MCP server (references/tool-mcp.md) with tools like slidev-update-slide and slidev-insert-slide, and supports a writable Monaco editor (references/editor-monaco-write.md) that saves changes directly to the filesystem.
Sanitization: No explicit sanitization or validation steps are documented for content imported from external files or URLs.
[DYNAMIC_EXECUTION]: Slidev is built on Vite and Vue, allowing for the inclusion of Vue components and scoped CSS within markdown slides. It also supports the Monaco editor for live coding and code execution (references/editor-monaco-run.md), which involves runtime execution of slide content.
[COMMAND_EXECUTION]: The documentation lists several CLI commands for development, building, and exporting presentations, such as slidev build, slidev export, and slidev --remote (references/core-cli.md).
[EXTERNAL_DOWNLOADS]: The skill mentions dependencies such as playwright-chromium for PDF export and various @iconify-json/* packages for icons. These are standard, well-known resources for the Slidev ecosystem.

slidev