The Agent Skills Directory

[SAFE]: The skill provides a set of markdown-based instructions and a Python script (scripts/audit.py) for design auditing. The script uses only standard Python libraries (re, sys, pathlib) and performs read-only regex searches on local files specified by the user. No suspicious subprocess calls or environment modifications were detected.
[DATA_EXFILTRATION]: No network operations or data transmission patterns were found in the provided code. The skill's instructions suggest using standard platform tools (e.g., a browser) to research design references on well-known, trusted platforms like Pinterest, Dribbble, and Siteinspire.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and analyze user-provided UI code (HTML/CSS/JS). While this constitutes an ingestion point for untrusted data, the skill includes explicit instructions for human oversight ("Humano decide, IA executa") and focuses on structural design improvements rather than executable logic. The attack surface is inherent to its primary purpose as an auditing tool and does not present an exploitable risk in its current configuration.

ui-humanizer