agent-architecture-audit
Finds bugs and problems in AI agent code across 12 layers, then gives you fixes ranked by importance.
Installation
Paste this into Claude Code, Cursor, or any agent that can run commands.
What this skill does
What it does: - Checks all 12 layers of an agent system for hidden problems like broken tool calls, old memory leaking into new conversations, and silent repair loops - Finds where the model is giving wrong answers because of wrapper code, not because the model itself is broken - Catches problems in prompts, memory, tool definitions, and output formatting When to use it: - Before releasing an agent or AI-powered feature to real users - When an agent worked yesterday but is broken today - When the same model works in a playground but fails in your code - When you have been debugging for 15 minutes without finding the root cause - When users say the agent is getting worse or tools are unreliable
The full skill text could not be loaded right now. It installs fine with the command above, or read it at the source.
Mirrored from the author's public source. Install counts from the open skills registry.