The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill generates presentation source code (HTML and React) from user-provided talk details, creating a vulnerability surface where malicious user input could be interpolated into the resulting deck. This risk is inherent to the skill's primary purpose of code generation.
Ingestion points: User prompts used to generate slide content as described in SKILL.md.
Boundary markers: Not explicitly used in the templates.
Capability inventory: File system writing and local server execution via the provided scaffolding scripts.
Sanitization: The skill relies on standard agent behavior and does not specify custom sanitization for generated content.
[EXTERNAL_DOWNLOADS]: The skill uses external resources from well-known and trusted services during scaffolding and at runtime.
The React track uses npm to install framer-motion, lucide-react, and tailwindcss from the official registry.
The HTML track includes a script tag fetching tailwindcss from cdn.tailwindcss.com.
Google Fonts are loaded from fonts.googleapis.com.
[COMMAND_EXECUTION]: The skill includes several utility scripts (scripts/new-html-deck.sh, scripts/new-react-deck.sh, scripts/add-remote.sh) that perform legitimate file system operations to scaffold and configure presentation projects.
[DYNAMIC_EXECUTION]: The skill dynamically generates executable code by populating HTML and React templates with user-specific data at runtime.

slidewright