The Agent Skills Directory

[DYNAMIC_EXECUTION]: The script scripts/design_system.py computes filesystem paths using user-provided input without adequate validation or sanitization. Specifically, the persist_design_system function constructs target directory paths using the project_name argument and individual file paths using the page argument. The sanitization logic (replace(' ', '-')) fails to remove or block parent directory references (..), which can lead to directory traversal. An attacker could potentially supply a crafted project or page name to cause the script to create folders or write Markdown files to arbitrary locations on the system where the agent has write permissions.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a vulnerability surface where untrusted input can manipulate filesystem operations.
Ingestion points: CLI arguments for query, project_name, and page in scripts/search.py used to drive the design system generation and persistence logic.
Boundary markers: There are no explicit instructions or delimiters provided to the agent to treat these inputs as untrusted or to ignore embedded control sequences during the persistence phase.
Capability inventory: The skill includes the capability to create nested directory structures (mkdir with parents=True) and write files (open in write mode) to the local filesystem.
Sanitization: The input sanitization is insufficient, as it only replaces space characters with hyphens and does not handle path traversal sequences like .. or special filesystem characters.

ui-ux-pro-max