The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by ingesting and processing untrusted user data.
Ingestion points: The skill reads user-provided HTML files from the references/templates/ directory and processes external content files (such as Markdown or text documents) during the slide generation process.
Boundary markers: The instructions specify ignoring text content in templates, but they lack explicit boundary markers or instructions to disregard potential commands embedded within the primary content source files.
Capability inventory: The skill generates multiple HTML files and can invoke an external /pptx skill for file conversion.
Sanitization: A strict prohibition of JavaScript in the generated output is enforced, which reduces execution risks, but the input is not explicitly sanitized against instruction-based overrides.
[EXTERNAL_DOWNLOADS]: The generated slides reference assets from well-known content delivery networks.
Resources: Tailwind CSS and Font Awesome are loaded from cdn.jsdelivr.net. Fonts are fetched from fonts.googleapis.com.
[EXTERNAL_DOWNLOADS]: The workflow references an external skill for PowerPoint conversion from a trusted repository.
Evidence: The instructions guide the user to install the /pptx skill from the anthropics GitHub organization if it is not already available in the session.

slidekit-create