The Agent Skills Directory

[COMMAND_EXECUTION]: The skill instructions direct the agent to execute a local Python search script (scripts/search.py) by interpolating user-provided requirements directly into command-line arguments. This creates a potential command injection vector if the user-supplied text is not correctly sanitized or quoted by the agent.
[COMMAND_EXECUTION]: The SKILL.md file includes instructions for environment setup that involve the use of sudo (e.g., sudo apt install python3), which requires executing commands with administrative privileges.
[EXTERNAL_DOWNLOADS]: The documentation provides instructions for installing Python 3 using standard platform package managers such as Homebrew for macOS, APT for Ubuntu, and WinGet for Windows.
[PROMPT_INJECTION]: The skill processes untrusted user input through its search logic and incorporates the retrieved design data into the agent's active context, representing an indirect prompt injection surface.
Ingestion points: The query parameter in scripts/search.py is the primary entry point for user-supplied strings.
Boundary markers: The output of the search tool is structured using Markdown headers and lists; however, the skill does not utilize specific delimiters or instructions to prevent the agent from obeying instructions potentially embedded within the data.
Capability inventory: The provided Python scripts are limited to reading local CSV files; no network communication or filesystem write capabilities were detected in the skill's code.
Sanitization: The core.py script implements a tokenization process that removes punctuation from search queries, providing a basic level of input sanitization.

ui-ux-pro-max