AI writes code faster than any reviewer can read it, so the review has to be mechanical. This series is about the configuration that makes an agent’s output answerable to something it cannot argue with: the compiler, three dozen linters, a cyclomatic complexity ceiling, a coverage gate, mutation testing, and a pre-commit hook that refuses the commit when any of it fails.

The name is the point. The leash is on the AI, not on the person holding it. The series starts with why Go’s constraints suit agent-written code, gives the complete project configuration with every setting explained, and ends with the gate that runs the whole thing before anything reaches a branch.

The same standards are packaged as a Claude Code skill, go-leash, which audits a Go project against them and scaffolds whatever is missing.