The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The design system persistence logic in scripts/design_system.py is susceptible to path traversal via unsafe interpolation of user-supplied data.
Ingestion points: User-provided strings (such as project names or queries) enter the system through the search.py CLI and are processed by DesignSystemGenerator.generate.
Boundary markers: Absent. The script does not implement delimiters or safety instructions to distinguish user-provided labels from system-level file paths.
Capability inventory: The persist_design_system function utilizes Path.mkdir(parents=True) and open(..., 'w') to create directories and write generated Markdown documentation to the local filesystem.
Sanitization: The script performs minimal sanitization by only replacing spaces with hyphens in the project_slug variable. It fails to filter or block directory traversal sequences (e.g., ../) or absolute path indicators, which could allow a malicious user to influence the agent to write files to arbitrary locations on the system where the agent is executing.
[EXTERNAL_DOWNLOADS]: The skill includes numerous references to external documentation and design assets.
Evidence: data/typography.csv contains links to Google Fonts, and various stack-specific files in data/stacks/ link to official developer documentation for React, Next.js, Svelte, and other common frameworks.
Context: These URLs point to well-known, established technology providers and are used solely for documentation and legitimate design asset referencing.

ui-ux-pro