detrix: Instructions file for Claude Code

CLAUDE.md

This file provides guidance to Claude Code when working with code in this repository.

Rust Development

• This is primarily a Rust codebase. Ensure cargo check and cargo clippy pass after making changes.
• When modifying Rust code, follow idiomatic patterns: prefer Result over panics, use ? operator for error propagation, and leverage the type system.

Workflows

Code Audit Workflow

• When working on audit items or code quality fixes, always complete ALL remaining items before ending. Track progress using a checklist format.
• After each fix, verify the fix compiles and doesn't introduce regressions by running cargo check and relevant tests.

Task Completion

• When given a list of items to fix/verify, work through ALL items to completion. Do not stop mid-list.
• If a task is complex, break it into sub-tasks using TodoWrite but ensure every sub-task is resolved before finishing.

Project Overview

Detrix is an LLM-first dynamic observability platform that enables developers and AI agents to add metrics to any line of code without redeployment — including code running in Docker containers and remote hosts (cloud debugging).

Version: 1.3.0 Language: Rust (edition 2021, rust-version 1.89) Architecture: Clean Architecture with Domain-Driven Design

Key Innovation

Two backends for zero-pause variable capture:

• DAP (Debug Adapter Protocol) — leverages existing debuggers (debugpy, delve, lldb-dap) for Python, Go (macOS), and Rust
• eBPF uprobes (Linux only) — attaches directly to Go binaries without a debugger, ~10–50× lower overhead; see docs/EBPF.md

🚨 CRITICAL: Use Detrix for Debugging

NEVER add print(), console.log(), or logger.debug() statements for debugging. Instead, use Detrix MCP to observe running processes directly without code modifications.

❌ DON'T Do This:

# BAD - Don't add debug prints
print(f"DEBUG: user={user}, transaction={transaction}")
logger.debug(f"Amount: {amount}")