The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The minimal-tiptap-setup.sh script and SKILL.md documentation include a command that fetches a component registry from a public GitHub repository (Aslam97/shadcn-minimal-tiptap) using the shadcn CLI tool. This is a standard practice within the shadcn ecosystem for component distribution.
[INDIRECT_PROMPT_INJECTION]: The editor templates are designed to ingest and render user-provided HTML or Markdown content. If an AI agent later processes the output of these editors, malicious instructions embedded in the rich text could potentially influence the agent's behavior.
Ingestion points: The content prop in the Editor component (templates/base-editor.tsx), the value prop in the MinimalTiptapEditor (templates/minimal-tiptap-setup.sh), and the setContent command in templates/common-extensions.ts.
Boundary markers: No explicit boundary delimiters are used in the provided templates to isolate the user-provided rich text content from potential instructions.
Capability inventory: The provided code is limited to client-side UI logic, React state management, and standard network requests (fetch) for image uploads; it does not contain dangerous execution capabilities like eval() or system-level access.
Sanitization: The skill relies on Tiptap's internal ProseMirror-based schema for content sanitization, which filters unknown tags and attributes, but it does not provide additional custom sanitization logic.

tiptap