Agent Architecture Analysis

If you're building LLM agents and want to know whether your architecture is actually production ready, this performs a deep 12 Factor Agents audit against your codebase. It searches for specific patterns across 13 factors like structured outputs, prompt ownership, state management, and human in the loop tooling, then scores each as Strong, Partial, or Weak with file level evidence and code citations. The workflow enforces hard gates so you get real findings, not vibes. Use it when evaluating a new agent system, auditing something before production, or planning architectural improvements. Think of it as a systematic code review specifically for agentic patterns, not a quick checklist.

Install to Claude Code

npx -y skills add existential-birds/beagle --skill agent-architecture-analysis --agent claude-code