The Agent Skills Directory

[SAFE]: The skill implements a professional software architecture workflow divided into grounding, sketching, agreement, and implementation phases. All operations are consistent with the stated purpose of a design-focused assistant.
[SAFE]: No instances of prompt injection, data exfiltration, or unauthorized privilege escalation were found. The skill operates within the local project context and utilizes internal platform protocols (skill://) for task delegation.
[SAFE]: The 'Sketch' phase explicitly implements security best practices for multi-agent systems by ensuring that candidate design runners are independent by construction, with separate sessions and output locations.
[SAFE]: Instructions provided to sub-agents (runners) in references/runner-prompt.md include constraints to prevent them from calling external tasks or starting further sub-agents, maintaining a controlled execution environment.
[SAFE]: No obfuscated content, hardcoded credentials, or suspicious remote dependencies were detected in the skill files or documentation.

architect