The Agent Skills Directory

[Persistence & File System Access] (MEDIUM): The scripts/search.py script allows users to persist design systems to the local filesystem using the --persist, --project-name, and --page arguments.
Evidence: In scripts/search.py, the project_slug and page_filename are generated using a simple .replace(' ', '-') on user-provided strings.
Impact: This is insufficient to prevent path traversal. An attacker could provide a project name like ../../etc to attempt writing files outside the intended design-system/ directory.
[Indirect Prompt Injection] (LOW): The skill ingests a user-provided query to generate a "complete design system recommendation" via the generate_design_system function.
Ingestion points: scripts/search.py via the positional query argument.
Boundary markers: None visible in the provided script wrapper; relies on downstream processing in design_system.py (not provided).
Capability inventory: The script has file-write capabilities (persist_design_system).
Sanitization: No sanitization of the search query is performed before passing it to the recommendation generator.
[Data Exposure] (SAFE): The CSV files contain UI/UX guidelines and code snippets. While some are labeled 'Critical' or 'High', these refer to accessibility and performance severity in the context of the design system, not security vulnerabilities within the skill's execution environment.

ui-ux-pro-max