The Agent Skills Directory

[DATA_EXPOSURE]: The skill uses local CSV files for its knowledge base. Analysis of all 28 files confirms no hardcoded credentials, API keys, or sensitive internal data are present.
[COMMAND_EXECUTION]: The skill executes local Python scripts (scripts/search.py, scripts/design_system.py) to process user requests. It does not spawn arbitrary shell commands or use dangerous functions like eval() or exec().
[DATA_EXFILTRATION]: No network operations (curl, requests, fetch) were detected in any of the logic scripts. The skill operates entirely on local data and user-provided queries.
[COMMAND_EXECUTION]: The persistence feature in scripts/design_system.py writes design documentation to a design-system/ directory. This is protected by _safe_slug() (input sanitization) and _assert_within() (path traversal prevention), ensuring files cannot be written to sensitive locations.
[INDIRECT_PROMPT_INJECTION]: The skill has a data ingestion surface via the search query. While it processes untrusted user input to filter its local database, it lacks the dangerous capabilities (like network access or administrative shell execution) that would make this surface critical. Findings indicate a standard surface risk with appropriate path-handling safeguards.

ui-ux-pro-max