The Agent Skills Directory

[SAFE]: The skill implements a robust design-first software engineering workflow. It leverages other internal agent skills (how, why, arena, interrogate) to perform grounding, multi-model design synthesis, and implementation. The use of sub-model runners in the arena phase is a standard pattern for high-fidelity reasoning and does not involve untrusted remote code execution or data exfiltration. No malicious patterns, obfuscation, or sensitive data access were found in the instructions or reference templates.

architect