The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The skill's prerequisite section includes instructions for the agent to use sudo for package installation (e.g., sudo apt update && sudo apt install python3). If the agent attempts to resolve environment issues by following these instructions, it could attempt to perform unauthorized administrative actions on the host system.
[COMMAND_EXECUTION]: The core workflow relies on executing a local Python script (.shared/ui-ux-pro-max/scripts/search.py) by constructing shell commands. The instructions direct the agent to interpolate extracted user keywords directly into these commands: python3 .shared/ui-ux-pro-max/scripts/search.py "<keyword>". This pattern is vulnerable to command injection if the user-provided keyword contains shell metacharacters (e.g., ;, &&, or backticks).
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted user input and uses it to drive critical tool execution without sufficient guardrails.
Ingestion points: User requirement strings (Product type, style keywords, industry) extracted from the initial prompt.
Boundary markers: Absent. The instructions do not provide delimiters or warnings to treat the extracted keywords as untrusted data.
Capability inventory: Shell command execution via python3 to run search scripts across multiple domains.
Sanitization: Absent. The skill lacks any instruction to validate, escape, or sanitize the user-provided keywords before they are used in shell commands.

ui-ux-pro-max