Working with coding agents
Practical guides from building Vakr and running agents on real work. Each one works without Vakr.
Tests first with coding agents: make the agent prove itA tests-first workflow for Claude Code and Codex: the agent writes failing tests, you approve them, it writes the code, and a hash check proves it.10 October 2026
How to run Claude Code overnight, safelyA working setup for unattended Claude Code runs: one git worktree per task, explicit permissions, a turn cap, and a short checklist for the morning.10 October 2026
How to review AI-written code without reading every lineA review order for agent-written code: approve the tests first, check what the agent claimed against exit codes, let a second model attack the diff.10 October 2026
How to manage multiple Claude Code and Codex sessionsRunning several coding agents at once is an attention problem. How to name sessions, get notified, keep one queue of questions, and know when to stop.10 October 2026
Git worktrees for AI coding agents: a practical guideThe git worktree commands you need to run coding agents side by side, a folder layout that stays tidy, cleanup, and the pitfalls nobody mentions first.10 October 2026
How to make your Claude Code weekly limit lastWhat uses up a Claude Code weekly limit, how to see where it went with /usage, and how to split it across projects so it lasts until the reset.10 October 2026