The Agent Skills Directory

[SAFE]: The skill is entirely instructional and does not contain malicious code or command execution patterns.- [EXTERNAL_DOWNLOADS]: Mentions documentation and reference links to repositories on GitHub and Google-hosted sites. These are well-known or author-owned sources provided for context and reference.- [DATA_EXFILTRATION]: Instructions specify reading project-related design files (e.g., DESIGN.md, .stitch/DESIGN.md) to inform UI generation. No access to sensitive user credentials or unauthorized data transmission is present.- [PROMPT_INJECTION]: Directs the agent to prioritize local design specifications over its own training-data defaults (referred to as "AI-slop"). This is a task-specific constraint and does not attempt to subvert safety guidelines or extract system instructions.

omd