agent-architect

The skill fragment functions as a comprehensive architectural planning scaffold for Claude Agent SDK-based autonomous systems. It is not inherently malicious, but its breadth—especially around runtime tool usage (Bash, pip), external MCP servers, dashboards, and skill packaging—creates meaningful supply-chain and runtime risks when implemented. Mitigation requires strict sandboxing, least-privilege tool access, explicit credential management, explicit data-flow boundaries, and governance for skill loading. Treat as elevated-risk until concrete, sandboxed implementations and secure-by-default configurations are in place.