agent-architecture-analysis

What This Skill Does

The agent-architecture-analysis skill provides a systematic framework for auditing AI-powered agentic systems against the 12-Factor Agents methodology. This skill parses your codebase to evaluate how your application handles structural concerns like natural language-to-tool conversion, prompt management, and context window optimization. By automating the search for specific patterns—such as Pydantic schema usage, Jinja2 template implementations, and custom message serialization—the skill identifies architectural drift and provides actionable compliance ratings (Strong, Partial, Weak) to help you improve system reliability and maintainability.

Installation

You can install this skill directly via the OpenClaw CLI using the following command: clawhub install openclaw/skills/skills/anderskev/agent-architecture-analysis

Use Cases

• Architectural Audits: Evaluate the robustness of an LLM-based system before a major release.
• Technical Debt Assessment: Identify hardcoded prompts and unvalidated model outputs that could cause runtime failures.
• Onboarding/Review: Quickly understand the structure of a new agentic codebase by seeing how it handles context and tool calling.
• CI/CD Integration: Incorporate agent compliance checks into your build pipeline to prevent the introduction of weak prompt management or unvalidated outputs.

Example Prompts

1. "Analyze my project in /home/user/code/customer-support-agent and tell me which factors are currently rated as Weak."
2. "Perform a 12-factor audit on the codebase at ./src and focus specifically on how we handle prompt versioning."
3. "Run a compliance check on my agent architecture and generate a summary of anti-patterns found in the prompts/ directory."

Tips & Limitations

• Tips: For best results, ensure your project follows standard Python naming conventions. Use the docs_path parameter if you have pre-existing architectural documentation to allow the agent to cross-reference code with design intent.
• Limitations: This skill performs static analysis based on grep patterns and heuristics. It cannot execute the code to observe runtime behavior. Complex dynamic prompt generation or highly obfuscated code may yield false negatives. Always review the detailed logs generated by the skill to verify compliance classifications.