The Agent Skills Directory

[SAFE]: The skill implements a structured meta-process for software design, focusing on grounding, sketching, and synthesis. It relies on standard internal delegation patterns (delegate_task via the arena skill) without introducing external dependencies or risky execution patterns.
[SAFE]: Task delegation is designed with isolation in mind, specifically stating that child agents receive only the task brief and not the full conversation context, which limits information exposure across agent boundaries.
[SAFE]: The skill includes explicit instructions for human-in-the-loop checkpoints (Phase C) and encourages adversarial pressure on designs through the 'interrogate' skill before implementation begins.
[SAFE]: No evidence of prompt injection, credential exfiltration, obfuscation, or unauthorized command execution was found in the skill or its referenced documentation files.

architect