The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script, scripts/lint_skeleton.py, to perform preflight checks and validation on HTML templates before they are published to the site.
[DATA_EXFILTRATION]: The skill is configured to record its usage by calling skills_tool(action="register_usage", skill_id="ucoz-template-creator-skill"). This transmits the skill's identity to a usage accounting service managed by the platform.
[INDIRECT_PROMPT_INJECTION]: The skill has a surface for indirect prompt injection because it ingests untrusted data from uCoz projects and has the authority to overwrite site templates.
Ingestion points: Site structure, existing templates, and design configurations are read from the ucoz-mcp server using actions like project_get and design_get (documented in references/FRAMEWORK.md).
Boundary markers: The skill uses specific HTML comments (e.g., , ) and platform-specific conditional syntax (<?if(...)?>) to define and separate structural blocks.
Capability inventory: The skill uses ucoz-mcp tools to perform site-wide modifications, including templates_tool.update_template (replacing the site's entire CSS) and skeleton_publish (overwriting all system templates).
Sanitization: The skill mitigates risks through local linting via scripts/lint_skeleton.py and mandatory server-side validation using tmaker_tool(action="skeleton_validate") before any write operations.

ucoz-template-creator-skill