agent-architecture-audit - Agent Architecture Audit – Full-Stack LLM Diagnostic
Diagnoses agent and LLM application failures across a 12-layer stack, producing severity-ranked findings with code-first fixes.
Tags
Updated: 2026-06-30Capabilities
Typical Inputs
Typical Outputs
What this skill does
- audit agent architecture layers
- search codebase for anti-patterns
- map failures to source layers
- rank findings by severity
- generate structured audit report
- produce ordered fix plan
Inputs
- Agent application source code
- Session logs and traces
- Prompt templates and tool schemas
- Memory files and knowledge bases
- Provider configuration settings
Outputs
- Severity-ranked audit findings report
- Architecture diagnosis with layer mapping
- Ordered code-first fix plan
Requirements
- Access to agent application source code
- ripgrep (rg) installed for code search
- Read, Write, Edit, Bash, Grep, Glob tools