The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The scripts/source_resource_snapshot.py utility fetches JSON-formatted design tokens and palettes from ramps.studio and springs.studio. The network operations are restricted to these hardcoded domains, and the script includes explicit warnings and logic to prevent the inclusion of credentials in query strings.
[COMMAND_EXECUTION]: Several files in scripts/tests/ utilize subprocess.run to execute the skill's internal Python scripts, such as create_plan.py and validate_plan.py, for automated testing. These executions are limited to the skill's own distribution and are used for structure verification.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process user-provided product briefs. It mitigates the risk of adversarial instructions within these briefs by enforcing a mandatory 'Review Gate' workflow, requiring explicit user approval of a versioned plan before any implementation steps occur.
[DYNAMIC_EXECUTION]: While the skill references the potential use of external frameworks like Remotion for video rendering, the references/plan-contract.md specifically prohibits the automatic installation, previewing, or rendering of such assets without explicit user intervention and separate risk assessment, maintaining a clear boundary against unauthorized execution.

uiux-design-orchestrator