The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user documents, outlines, or notes to generate self-contained HTML presentations. This creates a potential surface for indirect prompt injection if the source material contains malicious instructions or scripts intended to be rendered in the output file.
Ingestion points: Processes user-supplied content as the source for presentation slides as described in SKILL.md.
Boundary markers: The instructions lack specific boundary markers or requirements to ignore embedded instructions within the source document.
Capability inventory: The skill instructs the agent to write a complete HTML file (presentation.html) containing JavaScript and CSS templates.
Sanitization: There are no explicit instructions for the agent to sanitize or escape user-provided text before inserting it into the HTML slide templates defined in references/TEMPLATES.md.
[EXTERNAL_DOWNLOADS]: The generated presentations incorporate scripts and styles from well-known external sources and trusted organizations.
Evidence: References Reveal.js assets from cdn.jsdelivr.net, Lucide icons from unpkg.com, and fonts from fonts.googleapis.com within SKILL.md and references/NAVIGATION.md.

html-presentation