The Agent Skills Directory

[SAFE]: The skill consists entirely of Markdown-based documentation and reference material for the Slidev presentation framework. No executable scripts, obfuscated content, or malicious instructions were detected.
[COMMAND_EXECUTION]: The skill documents legitimate CLI commands for Slidev development, building, and exporting (e.g., slidev dev, slidev build). These are standard operational instructions for the tool.
[EXTERNAL_DOWNLOADS]: The documentation references official dependencies such as playwright-chromium for PDF exports and various @iconify-json packages for icons. These are well-known, standard dependencies within the Slidev ecosystem.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent on how to process Markdown files and external code snippets. While this provides a surface for indirect injection from untrusted presentation source files, the instructions are focused on standard slide building and do not include patterns to bypass agent safety guidelines.
Ingestion points: Markdown source files (slides.md), external snippets (<<<), and imported slides (src: ./other.md).
Boundary markers: Uses standard Markdown slide separators (---) and fenced code blocks.
Capability inventory: Includes shell command execution for builds, local filesystem writing via the monaco-write feature, and network exposure via the --tunnel flag.
Sanitization: Relies on the default parsing and rendering logic of the Slidev framework.
[DATA_EXFILTRATION]: Documents the --tunnel feature which utilizes Cloudflare Quick Tunnels to expose a local presentation to the internet. This is a legitimate, documented feature of Slidev intended for remote presentations.

slidev