The Agent Skills Directory

Unverifiable Dependencies & Remote Code Execution (MEDIUM): The script scripts/search.py imports functionality from core.py and design_system.py. These core logic files are missing from the skill package, preventing a safety review of the search implementation and the persistence logic. This obscures potential malicious behavior or insecure coding practices.
Indirect Prompt Injection (LOW): The skill possesses a significant surface for indirect prompt injection via the data it processes.
Ingestion points: scripts/search.py reads data from four CSV files: charts.csv, colors.csv, web-interface.csv, and jetpack-compose.csv.
Boundary markers: Absent. The search results are returned as raw markdown to the agent's context without delimiters or instructions to ignore embedded instructions found in the data.
Capability inventory: The skill has file-system write capabilities (via the --persist flag) and the ability to process user queries against external data.
Sanitization: No sanitization or escaping of the CSV content is performed in the provided wrapper script before it is returned to the agent.
Data Exposure & Exfiltration (MEDIUM): The --persist and --output-dir arguments in scripts/search.py allow the agent to write data to the filesystem. Without the source code for the persist_design_system function, it is impossible to verify if the skill prevents path traversal or the overwriting of sensitive system files via user-controlled paths.

ui-ux-pro-max