The Agent Skills Directory

[PRIVILEGE_ESCALATION]: The SKILL.md file includes prerequisites instructions for the user to install Python using sudo on Linux systems (sudo apt update && sudo apt install python3). While standard for software installation, instructions involving elevated privileges are an area of concern for AI agents that might attempt to execute them.
[COMMAND_EXECUTION]: The skill's primary functionality is driven by the execution of local Python scripts, specifically python3 skills/ui-ux-pro-max/scripts/search.py. The agent is instructed to run these commands repeatedly with user-supplied keywords, which represents a script execution surface.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection because it reads from local CSV data files and interpolates the results directly into the agent's context for synthesis into design implementations.
Ingestion points: Data is ingested from multiple CSV files in the data/ and data/stacks/ directories via the scripts/core.py module.
Boundary markers: The format_output function in scripts/search.py does not use specific delimiters or protective warnings to isolate retrieved data from system instructions.
Capability inventory: The skill performs file reads and executes local Python scripts. The instructions in SKILL.md encourage the agent to 'Synthesize all search results and implement the design,' which could lead to the generation of code influenced by the ingested data.
Sanitization: The BM25.tokenize method in scripts/core.py provides basic sanitization by removing punctuation and converting text to lowercase before processing search queries.

ui-ux-pro-max