The Agent Skills Directory

[SAFE]: The skill enforces strict platform safety rules, specifically the non-negotiable requirement to keep the uCoz platform copyright ($POWERED_BY$) visible. It also mandates user confirmation for all high-risk actions such as deleting files, uninstalling modules, or restoring backups.
[SAFE]: External resources such as Heroicons, Lucide, Tabler, and imagery services like Unsplash or Pexels are well-known technology and design services and are used appropriately for UI and content enhancement.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a broad attack surface as it regularly ingests and modifies untrusted data in the form of site templates and files.
Ingestion points: Site templates and files are read into the agent's context using tools like read_template, files_read, and page_get (documented in SKILL.md and ARCHITECTURE.md).
Boundary markers: The instructions lack explicit requirements for the agent to use delimiters or specific safety instructions when processing the content of existing templates, which could contain malicious injection payloads.
Capability inventory: The skill has extensive write capabilities including update_template, files_write, module_install, and page_update across various modules (SHOP.md, BLOG.md, USERS.md).
Sanitization: While the template language reference (LANGUAGE.md) provides a $JSENCODE$ function for safe JavaScript output, there are no general sanitization or validation requirements for instructions parsed from external site data.

ucoz-design-editor-skill