The Agent Skills Directory

[COMMAND_EXECUTION]: Instructions in SKILL.md recommend using sudo for environment setup, which represents a privilege escalation risk in an automated environment. Evidence: sudo apt update && sudo apt install python3.\n- [COMMAND_EXECUTION]: The skill relies on executing a local Python script (search.py) with arguments constructed from user-provided queries. This creates a surface for command injection if the arguments are not properly sanitized before shell execution. Evidence: python3 skills/ui-ux-pro-max/scripts/search.py "<product_type> <industry> <keywords>" --design-system.\n- [PROMPT_INJECTION]: The 'Master + Overrides' pattern writes instructions to local files and retrieves them later as authoritative rules, creating an indirect prompt injection surface.\n
Ingestion points: design-system/MASTER.md and design-system/pages/.\n
Boundary markers: Absent; the agent is instructed to 'prioritize its rules' from these files without explicit verification or isolation.\n
Capability inventory: The skill uses search.py to write these files and later reads them back into the agent context to influence code generation.\n
Sanitization: No sanitization or validation of the search result content is described before it is persisted as a source of truth.\n- [NO_CODE]: The skill references script and data directories via relative filesystem paths (../../../src/...) pointing outside the skill package rather than including the source code directly in the distributed folder.

ui-ux-pro-max