The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary workflow involves reading user-supplied HTML and transforming it by injecting CSS from assets/base.css and functional JavaScript snippets. This processing of untrusted external data without explicit sanitization instructions presents an injection surface.
Ingestion points: The skill ingests plain/unstyled HTML from the user as described in the SKILL.md workflow.
Boundary markers: There are no specific boundary markers or instructions to treat user-supplied HTML as untrusted data or to separate it from the injected code.
Capability inventory: The skill instructs the agent to inject <style> and <script> blocks. The provided JavaScript includes interactions with localStorage (via saveDraft) and the system clipboard (via copyToClipboard).
Sanitization: The skill lacks instructions for sanitizing, escaping, or validating the input HTML before applying styles or adding interactive elements.

html-style