The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: Retrieves design standards from a public repository on GitHub (vercel-labs) and brand configuration assets from the author's official community registry (iamstarchild.com).
[COMMAND_EXECUTION]: The design system persistence logic in scripts/design_system.py constructs file paths using unvalidated user input.
Evidence: The script uses the project_name parameter directly to create directories and files (design-system/<project_slug>/MASTER.md) using mkdir(parents=True) and open().
Risk: This lack of sanitization allows for potential path traversal, where files could be written outside the intended directory if malicious input is provided. The risk is mitigated by an enforced .md extension.
[INDIRECT_PROMPT_INJECTION]: The skill exposes an attack surface by ingesting external data from the web (brand assets and UX rules) which is then processed to provide instructions for the agent.
Ingestion points: SKILL.md (Brand Registry) and references/web-interface-guidelines-source.md (GitHub guidelines).
Boundary markers: Not explicitly implemented in the retrieval logic.
Capability inventory: The skill has file system write capabilities and network read capabilities.
Sanitization: Data is processed as configuration or text without specific filtering.

@349/starchild-design-pack