The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external sources (uCoz site templates, project designs, and platform rules) which could theoretically contain malicious instructions designed to influence the agent.
Ingestion points: Content is read through tools such as tmaker_read, templates_read, page_get, and design_get (SKILL.md).
Boundary markers: The framework uses structural HTML comments (e.g., ) to delimit service blocks, though it lacks explicit instructions for the model to disregard instructions embedded within the processed content.
Capability inventory: The skill possesses significant write and publication capabilities, including tmaker_write, tmaker_publish, and templates_write.update_template (SKILL.md).
Sanitization: The skill enforces a safety-first workflow including a mandatory local preflight check via a bundled Python script (scripts/lint_skeleton.py) and a server-side validation call (skeleton_validate) before any destructive publication occurs.
[COMMAND_EXECUTION]: The skill instructs the agent to run a local Python script for linting purposes.
Evidence: The instruction python scripts/lint_skeleton.py <framework.html> is used to verify the integrity of the generated code (SKILL.md).
Context: Analysis of the provided script (scripts/lint_skeleton.py) confirms it performs benign static analysis using regular expressions to ensure the presence of required platform variables and the absence of common formatting errors (mojibake).
[DYNAMIC_EXECUTION]: The skill generates and publishes website framework code and CSS stylesheets based on a design brief.
Evidence: The agent is responsible for composing HTML templates and complete CSS tables which are then published to the live uCoz site.
Control: This behavior is legitimate and central to the skill's purpose. It is governed by a strict validation process, user consent requirements, and automatic backup creation before modification.

ucoz-template-creator-skill