The Agent Skills Directory

[COMMAND_EXECUTION]: The skill's primary workflow involves the agent executing a local Python script scripts/search.py to query design databases. This is a functional requirement but constitutes command execution within the agent's environment.
[PRIVILEGE_ESCALATION]: The SKILL.md file includes setup instructions for Linux users to use sudo apt install python3, which requires administrative privileges to install dependencies.
[INDIRECT_PROMPT_INJECTION]: The skill implements a search mechanism that retrieves content from multiple CSV files in the data/ directory based on user-supplied keywords. This data is then used by the agent to generate UI code, creating a surface for indirect prompt injection.
Ingestion points: Data is read from files such as styles.csv, prompts.csv, and ux-guidelines.csv via the scripts/search.py tool.
Boundary markers: The output of the search script uses standard Markdown headers but lacks specific delimiters or instructions to treat the retrieved content as untrusted data.
Capability inventory: The agent is tasked with planning and implementing code (HTML, TSX, Vue, Svelte) and modifying projects based on the retrieved data.
Sanitization: The Python scripts (core.py and search.py) do not perform sanitization or filtering of the CSV content before presenting it to the LLM.

ui-ux-pro-max